Computational Intelligence and Security

Record linkage is a technique for linking records of different files that correspond to the same individual. Traditional record linkage methods needs that the files have some common variables to permit such link. In this paper we study the possibility of applying record linkage techniques when the files do not share variables. In this case we establish links based on structural information. For extracting this structural information we use in this paper twofold integrals.

Alert verification is a process which compares the information referred by an alert with the configuration and topology information of its target system in order to determine if the alert is relevant to its target system. It can reduce false positive alerts and irrelevant alerts. The paper presents an alert verification approach based on multi-level fuzzy comprehensive evaluation. It is effective in achieving false alert and irrelevant alerts reduction, which have been proved by our experiments. The algorithm can deal with the uncertainties better than other alert verification approaches. The relevance score vectors obtained from the algorithm facilitate the formulation of fine and flexible security policies, and further alert processing.

When automatically constructing large programs using program transformations, the number of possible transformations grows very fast. In this paper, we introduce and test a new way of combining several program transformations into one transformation, inspired by the combinatorial concept of Covering Arrays (CA).

We have equipped the ADATE automatic programming system with this new CA transformation and conducted a series of 18 experiments which show that the CA transformation is a highly useful supplement to the existing ADATE transformations.

Texture analysis is an important technique in many image processing areas, such as scene segmentation, object recognition, and shape&depth perception. However, most methods are restricted to issue of computational complexity and supervised problems. Accordingly, we propose a efficient method of segmenting texture that uses unsupervised learning schemes to discover a texture cluster without a pre-knowledge. This method applies 2D Gaussian filters to the clustered region iteratively, and the thresholding value for segmenting is automatically determined by analyzing histogram of the clustered inner-region. It can be acquired by the boundary tracing in the clustered region. In order to show the performance of the proposed method, we have attempted to build a various texture images, and the segmenting quality was measured according to the goodness based on the segmented shape of region. Our experimental results showed that the performance of the proposed method is very successful.

Generally speaking, dependency analysis based Bayesian network learning algorithms are of higher efficiency. J. Cheng’s algorithm is a representative of this kinds of algorithms, while its efficiency could be improved further. This paper presents an efficient Bayesian network learning algorithm, which is an improvement to J. Cheng’s algorithm that uses Mutual Information (MI) and Conditional Mutual Information (CMI) as Conditional Independence (CI) tests. Through redefining the equations for calculating MI and CMI, our algorithm could decrease a large number of basic operations such as logarithms, divisions etc. and reduce the times of access to datasets to the minimum. Moreover, to efficiently calculate CMI, an efficient method for finding an approximate minimum cut-set is proposed in our algorithm. Experimental results show that under the same accuracy, our algorithm is much more efficient than J. Cheng’s algorithm.

Finding common patterns is an important problem for several computer science subfields such as Machine Learning (ML) and Data Mining (DM). When we use graph-based representations, we need the Subgraph Isomorphism (SI) operation for finding common patterns. In this research we present a new approach to find a SI using a list code based representation without candidate generation. We implement a step by step expansion model with a width-depth search. The proposed approach is suitable to work with labeled and unlabeled graphs, with directed and undirected edges. Our experiments show a promising method to be used with scalable graph matching.

Self-poised ensemble learning is based on the idea of introducing an artificial innovation to the map to be predicted by each machine in the ensemble such that it compensates the error incurred by the previous one. We will show that this approach is equivalent to regularize the loss function used to train each machine with a penalty term which measures decorrelation with previous machines. Although the algorithm is competitive in practice, it is also observed that the innovations tend to generate an increasedly bad behavior of individual learners in time, damaging the ensemble performance. To avoid this, we propose to incorporate smoothing parameters which control the introduced level of innovation and can be characterized to avoid an explosive behavior of the algorithm. Our experimental results report the behavior of neural networks ensembles trained with the proposed algorithm in two real and well-known data sets.

We address the problem of building an integrated meta-level framework for time deliberation and parameter control for a system solving a set of hard problems. The trade-off is between the solution qualities achieved for individual problems and the global outcome under the given time-quality constraints. Each problem is modeled as an

anytime

optimization algorithm whose quality-time performance varies with different control parameter settings. We use the proposed meta-level strategy for generating a deliberation schedule and adaptive cooling mechanism for anytime simulated annealing (ASA) solving

hard

task sets. Results on task sets comprising of the traveling salesman problem (TSP) instances demonstrate the efficacy of the proposed control strategies.

Recently, actor-critic methods have drawn much interests in the area of reinforcement learning, and several algorithms have been studied along the line of the actor-critic strategy. This paper studies an actor-critic type algorithm utilizing the RLS(recursive least-squares) method, which is one of the most efficient techniques for adaptive signal processing, together with natural policy gradient. In the actor part of the studied algorithm, we follow the strategy of performing parameter update via the natural gradient method, while in its update for the critic part, the recursive least-squares method is employed in order to make the parameter estimation for the value functions more efficient. The studied algorithm was applied to locomotion of a two-linked robot arm, and showed better performance compared to the conventional stochastic gradient ascent algorithm.

A new dynamic clustering method using multi-objective evolutionary algorithm is proposed. As opposed to the traditional static clustering algorithms, our method implements variable length chromosome which allows the algorithm to search for both optimal cluster center positions and cluster number. Thus the cluster number is optimized during run time dynamically instead of being pre-specified as a parameter. We also introduce two complementary objective functions–compactness and connectedness instead of one single objective. To optimize the two measures simultaneously, the NSGA-II, a highly efficient multi-objective evolutionary algorithm, is adapted for the clustering problem. The simultaneous optimization of these objectives improves the quality of the resulting clustering of problems with different data properties. At last, we apply our algorithm on several real data sets from the UCI machine learning repository and obtain good results.

We introduce a novel system of interconnected Self- Organizing Maps that can be used to build feedforward and recurrent networks of maps. Prime application of interconnected maps is in modelling systems that operate with multimodal data as for example in visual and auditory cortices and multimodal association areas in cortex. A detailed example of animal categorization in which the feedworward network of self-organizing maps is employed is presented. In the example we operate with 18-dimensional data projected up on the 19-dimensional hyper-sphere so that the “dot-product” learning law can be used. One potential benefit of the multimodal map is that it allows a rich structure of parallel unimodal processing with many maps involved, followed by convergence into multimodal maps. More complex stimuli can therefore be processed without a growing map size.

A decision fusion based method is proposed to improve unsupervised image segmentation. After the step of cluster label adjustment, each kind of texture is fixed with the same label. Then three simple fusion operators are applied according to the knowledge of multi-classifier fusion. Compared with feature fusion, decision fusion can combine the advantages of different features more intuitively and heuristically. Experimental results on textures and synthetic aperture radar (SAR) image demonstrate its superiority over feature fusion on removing the impact of noise feature and preserving the detail.

We present a new method that improves the performance of the speech recognition system with the speaker-independent (SI) phonetic recognizer. The performance of the speech recognition system with the SI phonetic recognizer is worse than that of the speaker dependent system due to the mismatch between the training utterances and a set of SI models. A new training method that iteratively estimates the phonetic templates and transformation vectors is presented to reduce the mismatch using speaker adaptation techniques. The stochastic matching methods are used to estimate the transformation vectors for speaker adaptation. The experiment performed over actual telephone line shows that a reduction of about 40% in the error rates could be achieved as compared to the conventional method.

Quality of Service (QoS) requirements in networks with uncertain parameters has become a very important research issue in the areas of Internet, mobile networks and distributed systems. This is also a challenging and hard problem for the next generation Internet and mobile networks. It attracts the interests of many people. In this paper we propose a methodology to choose optimized fuzzy controller parameters using the genetic algorithms. Specifically, differentiated service scheme with feedback preference information (FPI) is studied in more detail to illustrate the implement of the new approach. Simulation shows that the approach is efficient, promising and applicable in ad hoc networks. The performance of this scheduler is studied using NS2 and evaluated in terms of quantitative metrics such as packet delivery ratio, average end-to-end delay. Simulation shows that the approach is efficient, promising and applicable in Diff-Serv networks.

This paper outlines an approach for applying acquiring numerical breast cancer image data and diagnosis using neural network algorithm in a way that is easy to classify between benign and malignance. This paper is an extended work related to our previous work [1]. In our previous work we used k-means algorithm to detect and diagnosis breast cancer tumor’s region. However, to find the better results from the algorithm we need to add more numerical parameters of the breast cancer image and this algorithm has limited usage when applied with more number of parameters. Even if the cancer tumor is abnormal it was quite difficult to distinguish among those tumors. This paper summarizes the different comparative study of neural network algorithms to get the best classification of the breast cancer and explains how to acquire more numerical parameters from the breast cancer image data, so that it can help doctors to diagnosis efficiently between benign and malignance tumors.

In this paper, based on the concept of Discretized Generalized Pursuit Algorithm (DGPA), the discretized generalized pursuit hierarchical structure learning algorithm is constructed which is applied to the hierarchical structure learning automata oprating in the P-model stationary random environment. The efficacy of our algorithm is demonstrated by the numerical simulation, in which the hierarchical structure learning automata is applied to the problem of the mobile robots going through an unknown maze (the maze passage problem of mobile robots).

Because the expert(s) usually give the judgment with an uncertainty degree in general decision-making problems, combined the analytic hierarchy process (AHP) with the basic theory of the triangular fuzzy number (TFN), a TFN-based AHP model is suggested. The proposed model makes decision-makers’ judgment more accordant with human thought mode and derives priorities from TFN-based judgment matrices regardless of their consistency. In addition, formulas of the model are normative, they can be operated by programming easily and no human intervention is needed while applying the model-based software system. The results of an illustrative case indicate that, by applying the proposed model, fair and reasonable conclusions are obtained and the deviation scope of the priority weight of every decision element is given easily.

This paper presents a tactics using fuzzy logic mediator that selects proper robot action depending on the positions and the roles of adjacent two robots. Conventional Q-learning algorithm, where the number of states increases exponentially with the number of robots, is not suitable for a robot soccer system, because it needs so much calculation that processing cannot be accomplished in real time. A modular Q-learning algorithm reduces a number of states by partitioning the concerned area, where mediator algorithm for cooperation of robots is used additionally. The proposed scheme not only reduces a number of a calculation but also combines a robot action selection with robot cooperation by means of fuzzy logic system.

A new reference trajectory of walking on the ground for a five-link biped robot, considering both the SSP and the DSP, is developed firstly. And a fuzzy wavelet neural network controller to generate walking gaits to follow the reference trajectories is presented subsequently. Furthermore, an error compensation algorithm is presented for high accuracy. The reference trajectories are designed by solving the coefficients of time polynomial functions of the trajectories of the hip and the swing tip, through the constraint equations. The FWN controller is trained as inverse kinematic model of the biped robot by backpropagation algorithm offline. Simulation results show that the FWN controller can generate the gaits following the reference trajectories as close as possible, and the error compensation algorithm can decrease the error rapidly by iterative calculation.

The regression is one of the fundamental problems in data mining, which is central to many applications of information technology. Various approaches have been presented for regression problem nowadays. However, many problems still exist, such as efficiency and model selection problem. This paper proposes a new approach to regression problem, visual regression problem (VRA) in order to resolve these problems. The core idea is to transfer the regression problem to classification problem based on Ancona theorem, which gives the mathematical equivalence between two problems; and then use visual classification approach, which is an efficient classification approach developed based on mimicking human sensation and perception principle, to solve the transformed classification problem and get an implicit regression function; and finally utilize some mathematical skills to obtain the explicit solution of the regression problem. We also provide a series of simulations to demonstrate that the proposed approach is not only effective but also efficient.

Naive Bayes is one of the most efficient and effective learning algorithms for machine learning, pattern recognition and data mining. But its conditional independence assumption is rarely true in real-world applications. We show that the independence assumption can be approximated by orthogonally rotational transformation of input space. During the transformation process, the continuous attributes are treated in different ways rather than simply applying discretization or assuming them to satisfy some standard probability distribution. Furthermore, the information from unlabeled instances can be naturally utilized to improve parameter estimation without considering the negative effect caused by missing class labels. The empirical results provide evidences to support our explanation.

In this paper, we firstly introduce particle swarm optimization to the problem of learning Bayesian networks and propose a novel structure learning algorithm using PSO. To search in DAG spaces efficiently, a discrete PSO algorithm especially for structure learning is proposed based on the characteristics of Bayesian networks. The results of experiments show that our PSO based algorithm is fast for convergence and could obtain better structures compared with GA based algorithms.

In this paper, we present an experimental study on applying a new dissimilarity measure to the

k

-modes clustering algorithm to improve its clustering accuracy. The measure is based on the idea that the similarity between a data object and cluster mode, is directly proportional to the sum of relative frequencies of the common values in mode. Experimental results on real life datasets show that, the modified algorithm is superior to the original

k

-modes algorithm with respect to clustering accuracy.

A new approach for protection of transmission line including TCSC is presented in this paper. The proposed method includes application of Fuzzy Neural Network for distance relaying of a transmission line operating with a thyristor controlled series capacitor (TCSC) protected by MOVs. Here the fuzzy neural network (FNN) is used for calculating fault location on the TCSC line. The FNN structure is seen as a neural network for training and the fuzzy viewpoint is utilized to gain insight into the system and to simplify the model. The number of rules is determined by the data itself and therefore, a smaller number of rules are produced. The network parameters are updated by Extended Kalman Filter (EKF) algorithm. with a pruning strategy to eliminate the redundant rules and fuzzification neurons resulting in a compact network structure . The input to the FNN are fundamental currents and voltages at the relay end, sequence components of current, system frequency and the firing angle with different operating conditions and the corresponding output is the location of the fault from the relaying point The location tasks of the relay are accomplished using different FNNs for different types of fault (L-G,LL-G,LL, LLL).

Computational intelligence has figured prominently in many solutions to the network intrusion detection problem since the 1990s. This prominence and popularity has continued in the contributions of the recent past. These contributions present the success and potential of computational intelligence in network intrusion detection systems for tasks such as feature selection, signature generation, anomaly detection, classification, and clustering. This paper reviews these contributions categorized in the sub-areas of soft computing, machine learning, artificial immune systems, and agent-based systems.

This paper deals with a switching PID controller using a genetic algorithm in a multi-nonlinear system. In controlling the nonlinear element of the system, there are some problems such as the limit cycle. In this study, a switching PID controller was proposed to solve problems caused by nonlinearities of system. The PID is a well-known robust controller. But, in a motor system case, it may have a limit cycle when proportional gains exceed limit. However, in other case, if the PID gain is relatively small, its torque characteristics can be too weak. In this case, the suggested switching PID controller was found to be a good approach for solving these problems despite there being difficulties in determining its boundary and gains at each boundary. In this paper, an improved genetic algorithm was used for identifying a motor system and to determining each gain for the controller. In particular, new type of crossover and mutation using a sigmoid function is applied to improve the searching ability based on the proposed improved genetic algorithm. All the processes are investigated through simulations and are verified experimentally in a real motor system.

In this paper, we propose a new constraint handling approach that transforms constrained optimization problem of any number of constraints into a two objective preference optimization problem. We design a new crossover operator based on uniform design methods ([8]), a new mutation operator using local search and preference, and a new selection operator based on the preference of the two objectives. The simulation results indicate the proposed algorithm is effective.

Differential Evolution (DE) is generally considered as a reliable, accurate, robust and fast optimization technique. DE has been successfully applied to solve a wide range of numerical optimization problems. However, the user is required to set the values of the control parameters of DE for each problem. Such parameter tuning is a time consuming task. In this paper, a self-adaptive DE (SDE) is proposed where parameter tuning is not required. The performance of SDE is investigated and compared with other versions of DE. The experiments conducted show that SDE outperformed the other DE versions in all the benchmark functions.

This paper presents a two-phase steady-state evolutionary algorithm (TSEA) for solving function optimization containing multiple global optima. The algorithm includes two phases: firstly,steady-state evolution algorithm is used to get sub-optimal solutions in the global search,it enables individual to draw closer to each optimal solution,thus population is divided into subpopulations automatically after the global search.Secondly,local search is carried in the neighborhood of the best individual of each subpopulation to obtain precise solutions. Comparing with other algorithms, it has the following advantages. (1) It designs a new multi-parent crossover operator with strong direction which can accelerate the convergence.(2) A novel replacement strategy is proposed to maintain the diversity of population.This strategy is very simple and effective with little computational cost.(3) Proposed algorithm needs no additional control parameter which depends on a special problem.The experiment results show that TSEA is very efficient for the optimization of multi-modal functions.

Since most public key cryptosystem primitives require the computation of modular exponentiation as their main building block, the problem of performing modular exponentiation efficiently has received considerable attention over the years. It is known that optimal (shortest) addition chains are the key mathematical concept for accomplishing modular exponentiations optimally. However, finding an optimal addition chain of length

r

is an

NP

-hard problem whose search space size is comparable to

r

!. In this contribution we explore the usage of a Genetic Algorithm (GA) approach for the problem of finding optimal (shortest) addition chains. We explain our GA strategy in detail reporting several promising experimental results that suggest that evolutionary algorithms may be a viable alternative to solve this illustrious problem in a quasi optimal fashion.

Evolvable hardware (EHW) has been employed in the circuit design automation domain, as an alternative to traditional human being designer. However, limited by the scalability of EHW, at present the scales of all the evolved circuits are smaller than the circuits designed by traditional method. In this paper, a character classification system for recognizing 16 characters was evolved by a novel evolution scheme: reconfigurable architecture-based intrinsic incremental evolution. The entire EHW system is implemented on one Xilinx Virtex xcv2000E FPGA that is fitted in the Celoxica RC1000 board. Hardware evolutionary result proved that the new method could bring us a scalable approach to EHW by efficiently limiting the chromosome string length and reducing the time complexity of evolutionary algorithm (EA).

This paper revisits the problem of Traffic Engineering (TE) to assess the relevance of using Gene Expression Programming (

GEP

) as a new fine-tuning algorithm in destination-based TE. We present a new TE scheme where link weights are computed using

GEP

and used as fine-tuning parameters in destination-based path selection. We apply the newly proposed TE scheme to compute the routing paths for the traffic offered to a 23- and 30-node test networks under different traffic conditions and differentiated services situations. We evaluate the performance achieved by the

GEP

algorithm compared to a memetic and the Open Shortest Path First (

OSPF

) algorithms in a simulated routing environment using the NS packet level simulator. Preliminary results reveal the relative efficiency of

GEP

compared to the memetic algorithm and

OSPF

routing.

Although genetic algorithm (GA) has the ability to do quick and stochastic global search, it can’t efficiently use the output information for systems. Ant algorithm (AA), on the other hand, is a parallel-proceed and distributive-forward system with a relatively slow speed for carrying out its solution. Incorporating GA and AA can improve their merits one for another. In this paper, the model and the method from the combination of GA and AA are proposed. The convergence of the method based on the Markov theory is analyzed. The experiment and analysis are conducted on the NP-hard problems for the cases of TSP30 (Travel Salesman Problem 30 cities) and CHN144 (China 144 cities). This work proves that the satisfactory solution sequence is monotonically decreasing and convergent. The results of simulations show that not only this combined algorithm is a step-by-step convergent process, but also its speed and effect of solving are quite satisfactory.

A new nonslicing floorplan representation, moving block sequence (MBS), is first proposed. The MBS is suitable for evolutionary algorithms since no extra constraints are exerted on the solution space. Furthermore, an organizational evolutionary algorithm is designed with the intrinsic properties of MBS in mind, called MBS-OEA. In experiments, 21 benchmarks from MCNC and GSRC are used to test the performance of MBS-OEA. The number of blocks in these benchmarks varies from 9 to 300. Comparisons are also made between MBS-OEA and some well-designed existing algorithms. The experimental results show that MBS-OEA can find high quality solutions even for the problems with 300 blocks. Therefore, MBS-OEA is competent for solving large scale problems.

In this paper, a hybrid genetic algorithm for solving constrained optimization problems is addressed. First, a real-coded genetic algorithm is presented. The simplified quadratic interpolation method is then integrated into the genetic algorithm to improve its local search ability and the accuracy of the minimum function value. Simulation results on 13 benchmark problems show that the proposed hybrid algorithm is able to avoid the premature convergence and find much better solutions with high speed compared to other existing algorithms.

Ensemble method has been deeply studied and widely used in the machine learning communities. Its basic idea can be represented as: A ‘weak’ learning algorithm that performs just slightly better than random guessing can be ‘boosted’ into an arbitrarily accurate ‘strong’ learning algorithm. Inspired from the fascinating idea, the paper used ensemble method to improve the performance of genetic algorithm and proposed an efficient hybrid optimization algorithm: GA ensemble. In GA ensemble, a collection of genetic algorithms are designed to solve the same problem and population of each algorithm is sampled from a solutions pool using bagging method. Experiments on combinatorial optimization problem and GA-deceptive problems show that ensemble method improves the performance of genetic algorithm greatly.

A parallel ant colony algorithm for mining the classification rules is presented. By preprocessing classification information on databases and combining ACO with parallel strategies, our algorithm could extract classification rules efficiently in parallel. Experimental results on several benchmark datasets show that our algorithm can discover classification rules more quickly with better accuracy, simplicity than other methods such as improved Ant-Miner algorithm and C4.5 based on well known decision tree algorithm.

Traditionally, product returns have been viewed as an unavoidable cost of distribution systems. Up to now there are few studies to address the problem of determining the number and location of centralized product return centers where returned products from retailers or end-customers are collected for manufacturers’ or distributors’ repair facilities while considering the distribution system. To fill the void in such a line of research, this paper proposes a nonlinear mixed-integer programming model and a genetic algorithm that can solve the distribution problem with forward and reverse logistics simultaneously. Compared with a partly enumeration method, the numerical analysis shows the effectiveness of the proposed model and its genetic algorithm approach.

To improve evolutionary design of circuits in efficiency, scalability and optimizing capability, a genetic algorithm based approach was proposed. It employs a gate-level encoding scheme supporting flexible changes of functions and interconnections of comprised logic cells, a multi-objective evaluation mechanism of fitness with weight-vector adaptation and circuit simulation, and an adaptation strategy for crossover probability and mutation probability to vary with individuals’ diversity and genetic-search process. It was validated by experiments on arithmetic circuits, obtaining circuits with expected functions, novel structures, and higher performances in gate usage and operating speed as compared with the results of both conventional and evolutionary approaches. Moreover, by studying the circuits evolved for problems of increasing scales, some novel, efficient and generalized principles have been discerned, which are easy to verify but difficult to dig out by human experts with existing knowledge.

The application of the auxiliary particle filter to the robot localization problem is considered. The auxiliary particle filter (APF) is an enhancement of the generic particle filter. However, APF suffers from the impoverishment problem and needs a large number of particles to represent the system posterior probability density function. An evolutionary computing method, the genetic algorithm is introduced into APF to remove early convergence yet improves the quality of potential solutions. Experiment results show that the evolutionary APF algorithm needs fewer particles and is more precise and robust for mobile robot localization in dynamic environment.

In this paper, we propose a programming model for the degree-constrained minimal bandwidth multicast routing problem in overlay networks, and design a hybrid genetic algorithm to solve it. The algorithm combines heuristic searching and genetic searching together for global optimal seeking. The complexity analysis and numerical experiments suggest our proposed model and algorithm are practical and efficient.

There are many non-probabilistic factors that affect the financial markets such that the returns of risky assets may be regarded as fuzzy numbers. This paper discusses the portfolio selection problem based on the possibilistic mean and variance of fuzzy numbers, which can better described an uncertain environment with vagueness and ambiguity to compare with conventional probabilistic mean-variance methodology. Markowitz’s mean-variance model is simplified a linear programming when returns of assets are symmetric triangular fuzzy numbers, so the possibilistic efficient portfolios can be easily obtained by some related algorithms.

The multi-criteria Minimum Spanning Tree problem is an NP-hard problem, and is difficult for the traditional network optimization techniques to deal with. In this paper, a novel genetic algorithm (NGA) is developed to deal with this problem. First, based on the topology of the problem, the proposed algorithm adopts a heuristic crossover operator and a new mutation operator. Then, in order to enhance the ability of exploration of crossover, a new local search operator is designed to improve the offspring of crossover. Furthermore, the convergence of the proposed algorithm to globally optimal solution with probability one is proved. The simulation results indicate that the proposed algorithm is effective.

Agent technology is widely used in the construction of large software systems, in particular E-Commerce and secure-critical systems. To fully utilize the potential of agents in the software system, it is essential to embed the BDI (

Beliefs

,

Desires

,

Intentions

) properties of agents in the software agents that model them. This paper introduces a formal software architectural design of a Multi-agent system (MAS) in which the BDI architecture is embedded. We embed the BDI properties of agents in an extended state machine (ESM) model and suggest that an implementation of the BDI architecture in a high-level programming language can be tested for conformance by generating test cases from the ESMs.

As the interest about Ubiquitous Computing has been increasing, it is actively processing research which advanced countries try to realize it such as Smart Space, Cool Town, Easy living, TRON project, and so on. The aim of these projects provides user oriented intelligent service considering relationship among main components (user, object, and environment) of a ubiquitous era. In this paper, we propose User-oriented intelligent Multimedia Service system in the Intelligent Home (IHUMS). The proposed system conducts intelligently the context information (user preference, user location, device status, etc.) using smart sensor agent module. It also provides the interoperability of multimedia among incompatible devices, authentication method which is suitable for the Intelligent Home, and transparent and secure service.

In this paper, we present an original meta-learning framework, namely the

Mindful

(Meta INDuctive neuro-FUzzy Learning) system.

Mindful

is based on a neuro-fuzzy learning strategy providing for the inductive processes applicable both to ordinary base-level tasks and to more general cross-task applications. The results of an ensemble of experimental sessions are detailed, proving the appropriateness of the system in managing meta-level contexts of learning.

This paper deals with cooperation for virtual reality applications. In a multi-agent system, cooperation between agents is an important element to solve a common task, which is very difficult or impossible for a single agent or a group of agents without cooperation. Hence we focus on cooperation in the predator-prey problem where a group of programmed and learning predators coordinates their actions to capture the prey. These actions of a learning predator are dynamically weighted by a behavioral system based on motor schemas and classifier systems. At each instant, the system must modify the weights in order to enhance the strategies of the group, as surrounding a prey. Thanks to the classifier system the learning predator learns situations and gradually adapts its actions to its environment. First encouraging results show that coupling such systems gives very efficient performances in dynamic environments.

Location-based services (LBS) depends on data gathered from mobile and ubiquitous devices and use to provide services like getting the appropriate location of a mobile user presented in physical and logical maps. The main operations of location management in LBS are updating and searching or paging. Some studies to improve these were presented by using optimal sequential paging and location area schemes. Different LBS means variety of methods on accessing data that leads to complexity of providing services. In this paper, we use an approach of hierarchical structured agents for the method of locating a mobile object in a location-based service. This study focuses on location management by using agents. Agents were used for accessing the distributed databases on LBS. It also introduces a hierarchical searching method that uses a nearest neighbor technique for fast searching. The result of using the technique shows an improved searching method in the location management.

This paper presents the principles and instrument structures of on_line measurement system for dielectric loss of virtual capacitive_type equipment. The system based on wavelets, LabVIEW, GPRS network and SPC correlation technics, which can efficiently solve the problem of eliminating noises from signals, removing the electromagnetism interference to measure the results of influence, global area and full auto on_line measurement, to attain the higher precision of measurement.

Model checking has being used mainly to check if a system satisfies the specifications expressed in temporal logic and people pay little attention to the problem for model checking logics of knowledge. However, in the distributed systems community, the desirable specifications of systems and protocols have been expressed widely in logics of knowledge. In this paper, based on the SMV, by the semantics of knowledge and set theory, approaches for model checking logics of knowledge and common knowledge are presented. These approaches make SMV’s functions extended from temporal logics to temporal logics of knowledge. We will illustrate in an example the applications to security verifications for a cryptographic protocol.

In this paper, we combine the syntax-based belief change approach and model-based approach, and present a computational approach for belief change. We introduce functions to revise or contract a belief set, as well as functions to revise or contract a belief base. We also show properties of the revision functions and the contraction functions. The implementation of the revision functions and the contraction functions are also considered, and algorithms to revise a belief set or contract a sentence from a belief set are also given. Compared with related works, the main characteristic of our approach is that the functions can be implemented by algorithms.

We present a feature selection method by fuzzy inference and its application to spam-mail filtering in this work. The proposed fuzzy inference method outperforms information gain and chi squared test methods as a feature selection method in terms of error rate. In the case of junk mails, since the mail body has little text information, it provides insufficient hints to distinguish spam mails from legitimate ones. To address this problem, we follow hyperlinks contained in the email body, fetch contents of a remote web page, and extract hints from both original email body and fetched web pages. A two-phase approach is applied to filter spam mails in which definite hint is used first, and then less definite textual information is used. In our experiment, the proposed two-phase method achieved an improvement of recall by 32.4% on the average over the 1

st

phase or the 2

nd

phase only works.

The paper begins with the in-depth dwelling of challenges posed by cellular network communication and further provides solution to those different challenges. A network that is adaptable to the demands of user is desired. For this reason it is necessary to provide agent based architecture, which is flexible, dynamic and can react proactively according to the services desired and demanded. The purpose of this research is to determine the best and an intelligent multi-agent framework for cellular networks by integrating various cooperative agents and wireless networks. Moreover the work justifies the agentification of the cellular network by discussing positive and negative aspects of the proposed feasible solution.

This paper first proposes a novel transitive dependence theory. Based on the proposed transitive dependence theory, a dynamic virtual organization formation framework has been worked out which includes service discovery, transitive dependence based reasoning for organization partners search and organization resolution.

Most current education organizations use books and CDs as the main media, which takes a long time for knowledge updating between education resource providers and the users. The rapid development of the Internet has brought with it the possibility of improving the resource purveying mechanisms. Therefore, we designed an agent based system to purvey education resources from the resource centre to schools through the Internet. Agent technology helps to improve system performance and flexibility. This paper describes the design of our system, details the functions of the main parts of the system, shows the communication methods between agents and finally evaluates the system by experiments.

Some actual multi-agent automated negotiation systems using auction mechanism in e-commerce are inefficient, as buyer negotiation agents are lack of enough rationality and uneasy to determine appropriate bid price automatically according to different circumstance. In order to improve the auction-based negotiation efficiency of multi-agent system in e-commerce, this paper proposes a model of game agent for auction-based negotiation and a bid negotiation algorithm based on game theory, which provide a new effective way to establish buyer game agent for making buyer agent determine price more accurately and bid more rationally. As buyer agents bid in a more rational way, auction agents and buyer agents can finish negotiation more efficiently.

In this paper, we design a novel autonomous mobile robot which uses quantum sensors to detect faint signals and fulfills some learning tasks using quantum reinforcement learning (QRL) algorithms. In this robot, a multi-sensor system is designed with SQUID sensor and quantum Hall sensor, where quantum sensors coexist with traditional sensors. A novel QRL algorithm is applied and a simple simulation example demonstrates its validity.

This paper focuses on the problem of cooperative search using a team of Unmanned Aerial vehicles (UAVs). The objective is to visit as many unknown area as possible, while avoiding collision. We present an approach which combines model predictive control(MPC) theory with genetic algorithm(GA) to solve this problem. First, the team of UAVs is modelled as a controlled system, and its next state is predicated by MPC theory. According to the predicted state, we then establish an optimization problem. By use of GA, we get the solution of the optimization problem and take it as the input of the controlled system. Simulation results demonstrate the feasibility of our algorithm.

This paper demonstrates the self-organization evolution of distributed Supply Networks (SNs) based on fitness landscape theory. The environment and the internal mechanism are the origin of SN evolution. The SN emerges from the local interaction of the firms to fulfill the stochastic demands. The collaboration among firms is path dependence. The evolution of a SN is self-reinforcement and sensitive to initial conditions, which may lead to multiple equilibrium state and chaos. The evolution result is non-deterministic and can not be predicted precisely. The long-term strategy is better than short-term strategy for a firm in SN collaboration to adapt to the environment.

Dynamic architecture of multi-agent systems (MAS) is very important for the critical systems. As the existing formal specifications cannot describe the dynamic architecture of MAS, while

π

-calculus is specially suited for the description and analysis of concurrent systems with dynamic or evolving topology, a formal approach using

π

-calculus is presented to describe MAS.

π

-calculus can describe the interactions among agents and permit their analysis for some key properties, e.g. deadlock, bisimilarity. By constructing a MAS model in electronic commerce, the modeling process using

π

-calculus are illustrated.

Cooperation between storage devices is an important aspect of autonomic storage system. By employing multiple distributed storage resources, storage system can greatly improve its performance. In this paper, agent-based methodology is introduced to build an autonomic storage system infrastructure. To select appropriate storage devices, queuing models are established to estimate the future storage device performance. A replica selection method and a data allocation algorithm are presented to gain an aggregate transfer rate according to the predicted performance. The results show that our models are useful for evaluating the mean response time of storage devices.

A new distributed spam filter system based on mobile agent is proposed in this paper. We introduce the application of mobile agent technology to the spam filter system. The system architecture, the work process, the pivotal technology of the distributed spam filter system based on mobile agent, and the Naive Bayesian filter method are described in detail. The experiment results indicate that the system can prevent spam emails effectively.

This paper presents the hexagon-based Q-leaning to find a hidden target object with multiple robots. We set up an experimental environment with three small mobile robots, obstacles, and a target object. The robots were out to search for a target object while navigating in a hallway where some obstacles were placed. In this experiment, we used two control algorithms: an area-based action making (ABAM) process to determine the next action of the robots and hexagon-based Q-learning to enhance the area-based action making process.

In this paper, we propose a naive statistics method for constructing a personalized recommendation system for the Electronic Program Guide (EPG). The idea is based on a primitive approach of N-gram to acquire nouns and compound nouns as prediction features, and then to combine the

$\it{tf\cdot idf}$

weighting to predict user favorite programs. Our approach unified feedback process, system can incrementally update the vector of extracted features and their scores. It was proved that our system has good accuracy and dynamically adaptive capability.

Text classification techniques mostly rely on single term analysis of the document data set, while more concepts especially the specific ones are usually conveyed by set of terms. To achieve more accurate text classifier, more informative feature including frequent co-occurring words in the same sentence and their weights are particularly important in such scenarios. In this paper, we propose a novel approach using sentential frequent itemset, a concept comes from association rule mining, for text classification, which views a sentence rather than a document as a transaction, and uses a variable precision rough set based method to evaluate each sentential frequent itemset’s contribution to the classification. Experiments over the Reuters corpus are carried out, which validate the practicability of the proposed system.

We examine an automated mechanism, which allows users to access this information in a structured manner by segmenting unformatted text records into structured elements, annotating these documents using XML tags and using specific query processing techniques. This research is the first step to make an automatic ontology generation system. Therefore, we focus on the explanation how we can automatically extract structure when seeded with a small number of training examples. We propose an approach based on Hidden Markov Models to build a powerful probabilistic model that corroborates multiple sources of information including, the sequence of elements, their length distribution, distinguishing words from the vocabulary and an optional external data dictionary. We introduce two different HMM models for information extraction from different sources such as bibliography and Call for Papers documents as a training dataset. The proposed HMM learn to distinguish the fields, and then extract title, authors, conference / journal names, etc. from the text.

Automatic text categorization is the task of assigning natural language text documents to predefined categories based on their context. In order to classify text documents, we must evaluate the values of words in documents. In previous research, the value of a word is commonly represented by the product of the term frequency and the inverted document frequency of the word, which is called

TF

*

IDF

for short. Since there is a different role for a word in different category documents, we should measure the value of the word according to various categories. In this paper, we proposal a new method used to measure the importance of words in categories and a new framework for text categorization. To verity the efficiency of our new method, we conduct experiments using three text collections. The k-NN is used as the classifier in our experiments. Experimental results show that our new method makes a significant improvement in all these text collections.

In 2003, Kim applied DCT technique to propose a content-based image copy detection method. He successfully detected the copies with or without modifications, and his method is the first that can detect the copies with water coloring and twirling modifications. However, Kim’s method can only detect copies modified with a 180 degree rotation. When copies are rotated by 90 or 270 degrees, Kim’s method fails to discover them. Also, his method can not deal with the copies with minor rotations like 1 degree or 5 degree rotation, and so on. To conquer this weakness, we apply ellipse track division strategy to extract the features and propose our methods. The experimental results confirm that our proposed method can successfully capture block features of an image even if it is rotated to any degree.

The explosive growth of information on the web demands effective intelligent search and filtering methods. Consequently, techniques have been developed that extract conceptual information to form a personalized view of the search context. In a similar vein, this system ventures to extract conceptual information as a weighted term category automatically monitoring the user’s browsing habits. This concept hierarchy can be served as a thematic search context to disambiguate the words in the user’s query to form an effective search query. Experimental results carried out with this framework suggests that implicit measurements of user interests, combined with the semantic knowledge embedded in concept hierarchy can be used effectively to infer the user context and to improve the results of information retrieval.

To achieve high quality of push-based information service, in this paper, collaborative filtering and content-based adaptability approaches are surveyed for user-centered personalized information, then based on the above method, we proposed a mixed two-phased recommendation algorithm for high-quality information recommendation, upon which performance evaluations showed that the mixed algorithm is more efficient than pure content-based or collaborative filtering methods, for pure of either approaches is not so efficient for the lack of enough information need information. And moreover we found with large amount registered users, it is necessary and important for the system to serve users in a group mode, which involved merged retrieval issues.

In this paper, an interest region-based image retrieval system (IRBIR) that combines similarity contributions from interest regions specified by user in images to form a single value for measuring similarity between images is proposed. The interest region-based framework utilizes the segmentation result to capture the higher-level concept of images. A novel image segmentation based on Graph-Cut is proposed for the final result. The segmentation method in this paper is fast and accurate enough for the real-time image retrieval demand than previous region-based methods. Experimental and comparison results, which are performed using a general purpose database containing 2,000 images, are encouraging.

The World Wide Web can be viewed as a gigantic distributed database including millions of interconnected hosts some of which publish information via web servers or peer-to-peer systems. We present here a novel method for the extraction of semantically rich information from the web in a fully automated fashion. We illustrate our approach via a proof-of-concept application which scrutinizes millions of web pages looking for clues as to the trend of the Chinese stock market. We present the outcomes of a 210-day long study which indicates a strong correlation between the information retrieved by our prototype and the actual market behavior.

Content-based information retrieval (CBIR) of multimedia data is an active research topic in intelligent information retrieval field. To support CBIR, high-dimensional data indexing and query is a challenging problem due to the inherent high dimension of multimedia data. Clustering-based indexing structures have been proved to be efficient for high-dimensional data indexing. However, most clustering-based indexing structures are static, in which new data cannot be inserted by just modifying the existing clusters or indexing structures. To resolve this problem, a two-level indexing method, called IASDS plus IPAT method, is developed in this paper. At the IASDS level, clusters and the corresponding subspaces can be incrementally updated, while the indexing structures within the clusters can be incrementally updated at the IPAT level. Furthermore, the proposed IASDS plus IPAT method is able to balance indexing efficiency and query accuracy by choosing an appropriate number of children nodes. The experimental results show that the IASDS plus IPAT method is very efficient for updating clusters and indexing structures with newly inserted data, and that its query accuracy is only slightly degraded while its query time is almost the same in comparison with the similar indexing structure built by non-incremental method.

Typhoon track prediction has mostly been achieved using numerical models which include a high degree of nonlinearity in the computer program. These numerical methods are not perfect and sometimes the forecasted tracks are far from those observed. Many statistical approaches have been utilized to compensate for these shortcomings in numerical modeling. In the present study, a support vector machine, which is well known to be a powerful artificial intelligent algorithm highly available for modeling nonlinear systems, is applied to predict typhoon tracks. In addition, a couple of input dimension reduction methods are also used to enhance the accuracy of the prediction system by eliminating irrelevant features from the input and to improve computational performance.

Support vector machines (SVMs) have been successfully applied to solve nonlinear regression and times series problems. However, the application of SVMs for tourist forecasting has not been widely explored. Furthermore, most SVM models are applied for solving univariate forecasting problems. Therefore, this investigation examines the feasibility of SVMs with backpropagation neural networks in forecasting tourism demand influenced by different factors. A numerical example from an existing study is used to demonstrate the performance of tourist forecasting. Experimental results indicate that the proposed model outperforms other approaches for forecasting tourism demand.

We present a new approach for modelling non-stationary time series, which combines multi-SVR and fuzzy segmentation. Following the idea of Janos Abonyi [11] where an algorithm of fuzzy segmentation was applied to time series, in this article we modify it and unite the segmentation and multi-SVR with a heuristic weighting on

ε

. Experimental results showing its practical viability are presented.

The design of new Space Launch Vehicle (SLV) involves a full set of disciplines – propulsion, structural sizing, aerodynamics, mission analysis, flight control, stages layout – with strong interaction between each other. Since multidisciplinary design optimization of multistage launch vehicles is a complex and computationally expensive. An efficient Least Square Support Vector Regression (LS-SVR) technique is used for trajectory simulation of multistage space launch vehicle. This newly formulation problem-about 17 parameters, linked to both the architecture and the command (trajectory optimization), 8 constraints – is solved through hybrid optimization algorithm using Particle Swarm Optimization (PSO) as global optimizer and Sequential Quadratic Programming (SQP) as local optimizer starting from the solution given by (PSO). The objective is to find minimum gross launch weight (GLW) and optimal trajectory during launch maneuvering phase for liquid fueled space launch vehicle (SLV).The computational cost incurred is compared for two cases of conceptual design involving exact trajectory simulation and with Least Square Support Vector Regression based trajectory simulation.

Transductive inference estimates classification function at samples within the test data using information from both the training and the test data set. In this paper, a new algorithm of transductive support vector machine is proposed to improve Joachims’ transductive SVM to handle various data distributions. Simulated annealing heuristic is used to solve the combinatorial optimization problem of TSVM, in order to avoid the problems of having to estimate the ratio of positive/negative samples and local optimum. The experimental result shows that TSVM-SA algorithm outperforms Joachims’ TSVM, especially when there is a significant deviation between the distribution of training and test data.

In this paper, an effective and simple method of input selection for nonlinear regression modeling using Support Vector Machine combined with Genetic Algorithm is proposed. Genetic Algorithm is used in order to extract dominant inputs from a large number of potential inputs in input selection process. Support Vector Machine is used as a nonlinear regressor with the selected dominant inputs. The proposed method is applied to the Box-Jenkins furnace benchmark to verify its effectiveness.

The paper proposed a hybrid two-stage method of support vector machines (SVM) to increase its performance in classification accuracy. In this model, a filtering stage of the

k

nearest neighbor (

k

NN) rule was employed to collect information from training observations and re-evaluate balance weights for the observations based on their influences. The balance weights changed the policy of the discrete class label. A novel idea of real-valued class labels for transferring the balance weights was therefore proposed. Embedded in the class label, the weights given as the penalties of the uncertain outliers in the classification were considered in the quadratic programming of SVM, and produced a different hyperplane with higher accuracy. The adoption of

k

NN rule in the filtering stage has the advantage to distinguish the uncertain outliers in an independent way. The results showed that the classification accuracy of the hybrid model was higher than that of the classical SVM.

Recently, multi-class SVMs have attracted much attention due to immense demands in real applications. Both the encoding and decoding strategies critically influence the effectiveness of the multi-class SVMs. In this work, a multi-class SVMs based on the SOM decoding algorithm is proposed. First, the binary SVM classifiers are trained according to the ECOC codes. Then the SOM network is trained with the output of the training samples and the optimum weights are obtained. Finally the unknown data is classified. By this method, the confidence of the binary classifiers is completely considered with the case avoided that the same minimum distance to several classes is obtained. The experimental results on the Yale face database demonstrate the superiority of the new algorithm over the widely-used Hamming decoding method.

XML has emerged as a medium for interoperability over the Internet. As the number of documents published in the form of XML is increasing there is a need for selective dissemination of XML documents based on user interests. In the proposed technique, a combination of Self Adaptive Migration Model Genetic Algorithm (SAMGA)[5] and multi class Support Vector Machine (SVM) are used to learn a user model. Based on the feedback from the users the system automatically adapts to the user’s preference and interests. The user model and a similarity metric are used for selective dissemination of a continuous stream of XML documents. Experimental evaluations performed over a wide range of XML documents indicate that the proposed approach significantly improves the performance of the selective dissemination task, with respect to accuracy and efficiency.

In this paper, we study a smoothing support vector machine (SVM) by using exact penalty function. First, we formulate the optimization problem of SVM as an unconstrained and nonsmooth optimization problem via exact penalty function. Second, we propose a two-order differentiable function to approximately smooth the exact penalty function, and get an unconstrained and smooth optimization problem. Third, by error analysis, we can get approximate solution of SVM by solving its approximately smooth penalty optimization problem without constraint. Compared with artificial neural network and time sequence, the precision of prediction of our smoothing SVM which is illustrated with the numerical experiment is better.

We propose a speech-act analysis method for Korean dialogue using Support Vector Machines (SVM). We use a lexical word, its part of speech (POS) tags, and bigrams of POS tags as utterance feature and the contexts of the previous utterance as context feature. We select informative features by

χ

2

statistic. After training SVMs with the selected features, SVM classifiers determine the speech-act of each utterance. In experiment, we acquired overall 90.5% of accuracy with dialogue corpus for hotel reservation domain.

Support Vector Machines (SVMs) for classification – in short SVM – have been shown to be promising classification tools in many real-world problems. How to effectively extend binary SVC to multi-class classification is still an on-going research issue. In this article, instead of solving quadratic programming (QP) in Algorithm

K

-SVCR and Algorithm

ν

-

K

-SVCR, a linear programming (LP) problem is introduced in our algorithm. This leads to a new algorithm for multi-class problem,

K

-class Linear programming

ν

–Support Vector Classification-Regression(Algorithm

ν

-

K

-LSVCR). Numerical experiments on artificial data sets and benchmark data sets show that the proposed method is comparable to Algorithm

K

-SVCR and Algorithm

ν

-

K

-SVCR in errors, while considerably faster than them.

This paper presents a new algorithm to deal with nominal attributes in Support Vector Classification by modifying the most popular approach. For a nominal attribute with

M

states, we translate it into

M

points in

M

– 1 dimensional space with flexible and adjustable position. Their final position is decided by minimizing the Leave-one-out error. This strategy overcomes the shortcoming in the most popular approach which assume that any two different attribute values have the same degree of dissimilarities. Preliminary experiments also show the superiority of our new algorithm.

In this paper, a new method that three-power spline function is used to smoothen the model of support vector machine (SVM) is presented. A third-order spline smooth support vector machine(TSSSVM) is obtained. Moreover, by analyzing the function precision, TSSSVM is better than SSVM and PSSVM.

This paper presents the comparableness of SVM method to artificial neural networks in the outlier detection problem of high dimensions. Experiments performed on real dataset show that the performance of this method is mostly superior to that of artificial neural networks. The proposed method, SVM served to exemplify that kernel-based learning algorithms can be employed as an efficient method for evaluating the revegetation potentiality of abandoned lands from coal mining activities.

T-cell activation is a pivotal process in immune response. A precondition for this activation is the recognition of antigenic epitopes by T-cell receptors. This recognition is antigen-specific. Therefore, identifying the pattern of a MHC restricted T-cell epitopes is of great importance for immunotherapies and vaccine design. In this paper, a new kernel is proposed to use together with support vector machine for the direct prediction of T-cell epitope. The experiment was carried on an MHC type I restricted T-cell clone LAU203-1.5. The results suggest that this approach is efficient and promising.

A Soft-magnetic ring cores (SMRC) inspection method using radial basis function support vector machine (RBFSVM) was developed. To gain the effective edge character of the SMRC, a sequence of image edge detection algorithms was developed. After edge was detected, feature vector was extracted. Subsequently, principal component analysis (PCA) is applied to reduce the dimension of the feature vector. Finally, RBFSVM is used for classification of SMRC, whose best accuracy in experiments is 97%.

In this paper, direct adaptive neural-network (NN) control is presented for a class of affine nonlinear systems in the strict-feedforward form with unknown nonlinearities. The states of the system can be stabilized. The control performance of the closed-loop system is guaranteed by suitably choosing the design parameters. Simulation results show the effectiveness of direct adaptive NN control.

This paper examines the performance of a recently proposed ACO algorithm when applied to the problem of constructing load-balanced clusters in ad hoc networks with node mobility. Performance, in this context, is measured in terms of the magnitude of change in solution quality after nodes move, and reactivity. Reactivity refers to the number of cycles ACO takes to recover from any degradation in solution quality resulting from node movements. Empirical results on 16 problem instances of various sizes revealed a positive correlation

The stochastic flow shop scheduling with uncertain processing time is a typical NP-hard combinatorial optimization problem and represents an important area in production scheduling, which is difficult because of inaccurate objective estimation, huge search space, and multiple local minima. As a novel evolutionary technique, particle swarm optimization (PSO) has gained much attention and wide applications for both function and combinatorial problems, but there is no research on PSO for stochastic scheduling cases. In this paper, a class of PSO approach with simulated annealing (SA) and hypothesis test (HT), namely PSOSAHT is proposed for stochastic flow shop scheduling with uncertain processing time with respect to the makespan criterion (i.e. minimizing the maximum completion time). Simulation results demonstrate the feasibility, effectiveness and robustness of the proposed hybrid algorithm. Meanwhile, the effects of noise magnitude and number of evaluation on searching performances are also investigated.

This paper presents a modified PSO algorithm, called the PSO with

C-Pg

mutation, or PSOWC-Pg, the algorithm adopts

C-Pg

mutation, the idea is to replace global optimal point

gBest

with disturbing point

C

and

gBest

alternately in the original formulae, the probability of using

C

is

R

. There are two methods for selecting

C

: stochastic method and the worst fitness method. The stochastic method selects some particle’s current position

x

or

pBest

as

C

stochastically in each iteration loop, the worst fitness method selects the worst particle’s

x

or the

pBest

of some particle with the worst fitness value as

C

. So, when

R

is small enough, the distance between

C

and

gBest

will tend towards 0, particle swarm will converge slowly and irregularly. The results of experiments show that PSOWC-Pg exhibit excellent performance for test functions.

Precise prediction of algal booms is beneficial to fisheries and environmental management since it enables the fish farmers to gain more ample time to take appropriate precautionary measures. Since a variety of existing water quality models involve exogenous input and different assumptions, artificial neural networks have the potential to be a cost-effective solution. However, in order to accomplish this goal successfully, usual problems and drawbacks in the training with gradient algorithms, i.e., slow convergence and easy entrapment in a local minimum, should be overcome first. This paper presents the application of a particle swarm optimization model for training perceptrons to forecast real-time algal bloom dynamics in Tolo Harbour of Hong Kong, with different lead times on the basis of several input hydrodynamic and/or water quality variables. It is shown that, when compared with the benchmark backward propagation algorithm, its results can be attained both more accurately and speedily.

The particle swarm optimization algorithm presents a new way for finding an optimal solution of complex optimization problems, where each particle represents a solution to the problem. In this paper a modified particle swarm optimization algorithm is applied to the optimization of the antenna array. Adding an item of integral control and the contractive factor in the modified algorithm can improve its global search ability. Simulation results show that the optimal pattern of the antenna array is able to approach the desired pattern. The results also demonstrate that the modified algorithm is superior to the original algorithm and the nonlinear least-square method.

This paper presents the application of an Ant Colony Optimization (ACO) algorithm approach for communications networks design problem. We explore the use of ACO’s for solving a network optimization problem, the degree-constrained minimum spanning tree problem (d-MST), which is a NP-Hard problem. The effectiveness of the proposed algorithm is demonstrated through two kinds of data set: structured hard (SHRD) complete graphs and misleading (M-graph) complete graphs. Empirical results show that ACO performs competitively with other approaches based on evolutionary algorithm (EA) on certain instance set problem.

To improve the linearly varying inertia weigh particle swarm optimization method (LPSO), a new concept of Crowd Avoidance is introduced in this paper. In this newly developed LPSO (CA-LPSO), particles can avoid entering into a crowded space while collaborate with other particles searching for optimum. Four well-known benchmarks were used to evaluate the performance of CA-LPSO in comparison with LPSO. The simulation results show that, although CA-LPSO falls behind LPSO when optimizing simple unimodal problems, it is more effective than LPSO for most complex functions. The crowd avoidance strategy enables the particles to explore more areas in the search space and thus decreases the chance of premature convergence.

This paper presents a new method for effectively searching all global minima of a multimodal function. The method is based on particle swarm optimizer, particles are dynamically divided into serval subgroups of different size in order to explore variable space using various step size simultaneously. In each subgroup, a new scheme is proposed to update the the positions of particles, this scheme takes into consideration the effect of all subgroup seeds. Experimental results for one dimensional, two dimensional and thirty dimensional test suites demonstrated that this method can get overall promising performance over a wide range problems.

This paper presents an outcome-space branch and bound algorithm for globally solving the linear multiplicative programming problems, in which we use a new two-level partition technique on rectangles and solve a simple relaxed quasiconcave programming problem at each branch node over outcome-space.We prove that the proposed algorithm is convergent. It can be shown by the numerical results that the proposed algorithm is effective.

This paper proposes a Binary Ant System (BAS), a binary version of the hyper-cube frame for Ant Colony Optimization applied to unconstrained function optimization problem. In BAS, artificial ants construct the solutions by selecting either 0 or 1 at every bit stochastically biased by the pheromone level. For ease of implementation, the pheromone value is designed specially to directly represent the probability of selection. Principal settings of the parameters are analyzed and some methods to escape local optima, such as local search and pheromone re-initialization are incorporated into the proposed algorithm. Experimental results show that the BAS is able to find very good results for the unconstrained function optimization problems of different characteristics.

We put forward a new conception,

dynamic association rule

, which can describe the regularities of changes over time in association rules. The dynamic association rule is different in that it contains not only a support and a confidence but also a

support vector

and a

confidence vector

. During the mining process, the data used for mining is divided into several parts according to certain time indicators, such as years, seasons and months, and a support vector and a confidence vector for each rule are generated which show the support and the confidence of the rule in each subsets of the data. By using the two vectors, we can not only find the information about the rules’ changes with time but also predict the tendencies of the rules, which ordinary association rules can not offer.

In the field of cluster analysis, most of existing algorithms are developed for small data sets, which cannot effectively process the large data sets encountered in data mining. Moreover, most clustering algorithms consider the contribution of each sample for classification uniformly. In fact, different samples should be of different contribution for clustering result. For this purpose, a novel typical-sample-weighted clustering algorithm is proposed for large data sets. By the atom clustering, the new algorithm extracts the typical samples to reduce the data amount. Then the extracted samples are weighted by their corresponding typicality and then clustered by the classical fuzzy

c

-means (FCM) algorithm. Finally, the Mahalanobis distance is employed to classify each original sample into obtained clusters. It is obvious that the novel algorithm can improve the speed and robustness of the traditional FCM algorithm. The experimental results with various test data sets illustrate the effectiveness of the proposed clustering algorithm.

This paper proposes the problem of mining weighted generalized fuzzy association rules with fuzzy taxonomies (WGF-ARs). It is an extension of the generalized fuzzy association rules with fuzzy taxonomies problem. In order to reflect the importance of different items, the notion of generalized weights is introduced, and leaf-node items and ancestor items are assigned generalized weights in our WGF-ARs. The definitions of weighted support and weighted confidence of WGF-ARs is also proposed. Then a new mining algorithm for WGF-ARs is also proposed, and several optimizations have been applied to reduce the computational complexity of the algorithm.

Different from familiar clustering objects, text documents have sparse data spaces. A common way of representing a document is as a bag of its component words, but the semantic relations between words are ignored. In this paper, we propose a novel document representation approach to strengthen the discriminative feature of document objects. We replace terms of documents with concepts in WordNet and construct a model named Concept CHain Model(CCHM) for document representation. CCHM is applied in both partitioning and agglomerative clustering analysis. Hierarchical clustering processes in different levels of concept chains. The experimental evaluation on textual data sets demonstrates the validity and efficiency of CCHM. The results of experiments with concept show the superiority of our approach in hierarchical clustering.

Time series are comprehensively appeared and developed in many applications. Similarity search under time warping has attracted much interest between the time series in the large databases. DTW (Dynamic Time Warping) is a robust distance measure and is superior to Euclidean distance. Nevertheless, it is more unfortunate that DTW has a quadratic time and the false dismissals are come forth since DTW distance does not satisfy the triangular inequality. In this paper, we propose an efficient range query algorithm based on a new similarity search method under time warping. When our range query applies for this method, it can remove the significant non-qualify time series as early as possible. Hence, it speeds up the calculation time and reduces the number of scanning the time series. Guaranteeing no false dismissals the lower bounding function is advised that consistently underestimate the DTW distance and satisfy the triangular inequality. Through the experimental results, our range query algorithm outperforms the existing others.

In this paper, an efficient and robust scene change detection algorithm is proposed by using low-level audiovisual features and several classification methods. The proposed algorithm consists of three stages. The first stage is shot boundary detection by using Support Vector Machine (SVM) and the second stage is the scene boundary detection using shot clustering based on visual information. In the last stage, the scene boundary correction with audio information is described.

Based on the FP-tree data structure, this paper presents an efficient algorithm for mining the complete set of positive correlated item pairs. Our experimental results on both synthetic and real world datasets show that, the performance of our algorithm is significantly better than that of the previously developed Taper algorithm over practical ranges of correlation threshold specifications.

As a simple, effective and nonparametric classification method, kNN algorithm is widely used in text classification. However, there is an obvious problem: when the density of training data is uneven it may decrease the precision of classification if we only consider the sequence of first k nearest neighbors but do not consider the differences of distances. To solve this problem, we adopt the theory of fuzzy sets, constructing a new membership function based on document similarities. A comparison between the proposed method and other existing kNN methods is made by experiments. The experimental results show that the algorithm based on the theory of fuzzy sets (fkNN) can promote the precision and recall of text categorization to a certain degree.

Traditional information retrieval technologies can satisfy users’ needs to some extent. But they cannot satisfy any query from different backgrounds, with different intentions and at different time because of their all-purpose characteristics. An integrated searching algorithm by combining filtering with collaborative technologies is presented in this paper. The user model is represented as the probability distribution over the domain classification model. A method of computing similarity and a method of revising user model are provided. Compared with the vector space model, the probability model is more effective on describing users’ interests. Furthermore, collaborative-based technologies are used, and as a result the scalability of the new algorithm is highly enhanced.

As a novel research direction, privacy-preserving data mining (PPDM) has received a great deal of attentions from more and more researchers, and a large number of PPDM algorithms use randomization distortion techniques to mask the data for preserving the privacy of sensitive data. In reality, for PPDM in the data sets, which consist of terabytes or even petabytes of data, efficiency is a paramount important consideration in addition to the requirements of privacy and accuracy. Recently, EMASK, an efficient privacy-preserving frequent pattern mining algorithm, was proposed. Motivated by EMASK, in this paper, we improve on it, and present an improved algorithm BV-EMASK to furthermore enhance efficiency. Performance evaluation shows that BV-EMASK reduces the execution time significantly when comparing with EMASK.

We present a cost model for predicting the performance of R-tree and its variants. Optimization base on the cost model can be apply in R-tree construction. we construct a new R-tree variant called CR*-tree using this optimization technique. Experiments have been carried out ,results show that relative error of the cost model is around 12.6%,and the performance for querying CR*-tree has been improved 4.25% by contrast with R*-tree’s.

In this paper, we will propose a novel outlier mining algorithm, called

Grid-ODF

, that takes into account both the local and global perspectives of outliers for effective detection. The notion ofOutlying Degree Factor

(ODF)

, that reflects the factors of both the density and distance, is introduced to rank outliers. A grid structure partitioning the data space is employed to enable Grid-ODF to be implemented efficiently. Experimental results show that Grid-ODF outperforms existing outlier detection algorithms such as LOF and KNN-distance in terms of effectiveness and efficiency.

It is important to perform the clustering task on XML documents. However, it is difficult to select the appropriate parameters’ value for the clustering algorithms. Meanwhile, current clustering algorithms lack the effective mechanism to detect outliers while treating outliers as “noise”. By integrating outlier detection with clustering, the paper takes a new approach for analyzing the XML documents by structure. After stating the concept of common neighbor based outlier, the paper proposes a new clustering algorithm, which stops clustering automatically by utilizing the outlier information and needs only one parameter, whose appropriate value range is decided in the outlier mining process. After discussing some features of the proposed algorithm, the paper adopts the XML dataset with different structure and other real-life datasets to compare it with other clustering algorithms.

The projection-type neural networks

$\tau \frac{dx}{dt}=-x+P_{\Omega }(x-\Lambda (t)\partial ^{0}E(x))$

are generic and useful models for solving the constrained optimization problems min {

E

(

x

)|

x

 ∈ Ω}. In the existing convergence/ stability analysis, the most are deduced based on the assumptions that

E

is uniformly or strictly convex and Ω is box-shaped. In this talk we present a generalized theory on convergence/stability of the networks. In the general setting that

E

is only convex and Ω is any closed bounded convex set, it is shown the global convergence/asymptotic stability of the networks in a specified sense. The presented theory sharpens and generalizes the existing results, and, consequently, underlies the applicability of the neural networks for a broader type of optimization problems.

In this paper, we proposed a novel hierarchical algorithm to recognize English calling cards. The algorithm processes multiresolution images of calling cards hierarchically to extract characters and recognize the characters by using an enhanced neural network method. Each processing step functions at lower overhead and results improved output. That is, first, horizontal smearing is applied to a 1/3 resolution image in order to extract the areas that only include characters from the calling card image. Second vertical smearing and the contour tracking masking, is applied to a 1/2 resolution image in order to extract individual characters from the character string areas. And last, the original image is used in the recognition step, because the image accurately includes the morphological information of the characters accurately. To recognize characters with diverse font types and sizes, the enhanced RBF network that improves the middle layer based on the ART1 was used. The results of experiments on a large number of calling card images showed that the proposed algorithm greatly improves the character extraction and recognition compared with traditional recognition algorithms.

Based on the concept of Immunodominance and Antibody Clonal Selection Theory, This paper proposes a new artificial immune system algorithm, Immune Dominance Clonal Multiobjective Algorithm (IDCMA), for multiobjective 0/1 knapsack problems. IDCMA divides the individual population into three sub-populations and adopts different evolution and selection strategies at them, but the update of each sub-population is not carried out all alone. The performance comparisons among IDCMA, SPEA, HLGA, NPGA, NSGA and VEGA show that IDCMA clearly outperforms the other five MOEAs in terms of solution quality.

This paper proposes a rate-distortion (RD) based seeded region growing (SRG) for extracting an object such as breast tumors in ultrasound volumes which contain speckle noise and indistinct edges. In the proposed algorithm, region growing proceeds in such a way that the growing cost is minimized which is represented as the combination of rate measuring the roughness of a region contour and distortion measuring the inhomogeneity of pixels in a region. An input image is first segmented into an initial seed region and atomic homogeneous regions. The seed is next merged with one of adjacent regions which makes the RD cost minimum at each step. Such a merging is repeated until the RD cost averaged over the entire seed contour reaches the maximum. As a result, the final seed holds region homogeneity and has a smooth contour while maximizing inhomogeneity against its adjacent regions. Experiments of extracting breast tumors in four real ultrasound volumes show the proposed method yields the average 40% improvement in error rate with respect to the results extracted manually over some conventional methods.

Microarray technology has been widely used in biological and medical research to observe a large number of gene expressions. However, such experiments are usually carried out with few replica or instances, which may lead to poor modelling and analysis. This paper suggests an approach to improve classification by using synthetic data. A new algorithm is proposed to estimate synthetic data value and the generated data are labelled by ensemble methods. Experiments with artificial data and real world data demonstrate that the proposed algorithm is able to generate synthetic data on uncertain regions of classifiers to improve effectiveness and efficiency of classification on microarray data sets.

A method to estimate long distance navigation of a mobile robot is proposed. The method uses the dead reckoning,sonar and infrared sensors to detect the landmarks. A corridor environment with equal spaced convex edges is applied as the mobile robot’s moving space and the convex edges are used as landmarks for the robot mounted with the combined sensor system to estimate its position. The robot detects the convex edges using combined sensor system, and navigates in this corridor by using the information obtained from dead reckoning and combined sensor system based on the Extended Kalman. Experiment result show the effectiveness of the method.

We present a novel algorithm SASAlignSimulated annealing with star-alignment. In the SASAlign, instead of starting with an initial solution chosen at random, we use the results formed by star-alignment to give a good starting point as the initial solution to the SA for further refinement. The time required by the algorithm scales linearly with the number of sequences in S, linearly with the number of iterations, and cube with the length of the sequences, that is

O

(

Nnl

3

). Experiments on the BAliBASE benchmark database also show that the proposed algorithm is efficient, and prove to be competitive with and better than the other method HMMT.

There are two important factors to visualize ultrasound datasets for volume ray casting method. The first is an efficient method to skip over empty space and the second is an adequate noise filtering method. We propose a noise-insensitive hierarchical min-max octree. In preprocessing stage, we generate a filtered dataset and make a hierarchical min-max octree from the dataset. In rendering step, we exploit the hierarchical min-max octree only when rays skip over transparent region. If rays reach meaningful object, color and opacity values are computed from the original volume dataset. By adaptively using two datasets, our method increases image quality while reducing rendering time.

In this paper, a novel fusing method for fundus retinal images based on robust registration techniques is proposed. In order to construct precise fusion map, we apply a ‘coarse-to-fine’ mapping strategy to accurately align pairs of identified vascular trees of retinas. A coarse mapping algorithm that exploits rigid model is first performed to maximize the goodness of fit between the vascular features over two time periods. However, the results suffer from local misalignment due to the inherent imprecise characteristics of the simplified model. A fine mapping algorithm is employed to eliminate ‘ghost vessels’ based on a local elastic matching technique. The transformed vectors for pixels in the registered fundus image are conveniently calculated by combining the local move vector and the global model transformed vector. Experiment results demonstrate nearly perfect fusion maps of several retinal fundus images in terms of visual inspection.

An Intelligent Particle Swarm Optimization (IPSO) for MO problems is proposed based on AER (Agent-Environment-Rules) model, in which Competition and Clonal Selection operator are designed to provide an appropriate selection pressure to propel the swarm population towards the Pareto-optimal front. Simulations and comparison with NSGA-II and MOPSO indicate that IPSO is highly competitive.

Based on the clonal selection theory, a new Dynamic Multiobjective Optimization (DMO) algorithm termed as Clonal Selection Algorithm for DMO (CSADMO) is presented. The clonal selection, the nonuniform mutation and the distance method are three main operators in the algorithm. CSADMO is designed for solving continuous DMO and is tested on two test problems. The simulation results show that CSADMO outperforms another Dynamic Evolutionary Multiobjective Optimization (EMO) Algorithm: a Direction-Based Method (DBM ) in terms of finding a diverse set of solutions and in converging near the true Pareto-optimal front (POF) in each time step.

Key frame extraction has been recognized as one of the important research issues in video retrieval. Key Frame Extraction based on Evolutionary Artificial Immune Network (KFE-EAIN) is proposed in this paper. To describe the distribution of video frame data, an artificial immune network is first evolved by video frame data. Then, key frame can be selected by minimal spanning tree of the network. KFE-EAIN does not require the number of clusters to be known beforehand. Otherwise, it can apply to both single shot and video sequence. Experimental results show that KFE-EAIN can effectively summarize content of a video with acceptable complexity.

Based on the Antibody Clonal Selection Theory of immunology, four immunologic regulation operators inspired by immune regulation mechanism of biology immune system are presented in this paper, and a corresponding algorithm, Immunologic Regulation Clonal Selection Algorithm (IRCSA), is put forward. The essential of immunologic regulation operators is to make fine adjustment among the candidates of the algorithm so as to make interrelations between antibodies more complicated and improve the stability, robustness and accuracy of the algorithm. Numeric experiments of function optimization indicate that the new algorithm is effective and useful.

In this paper, we propose a topology control algorithm for constructing an energy-efficient spanning subgraph for a wireless multi-hop network. The constructed topology has the following properties: (1) it preserves a minimum-energy path between every pair of nodes; (2) it is biconnected, i.e., it can tolerate any one node failure and avoid network partition. Simulation results show that the constructed topology has a small average node degree, a small average transmission range and a constant power stretch factor.

Vascular stent is a small tubular device expanded into stenotic artery to restore natural blood flow. This paper introduces application background and research overview of medical vascular stent. A tubular mini stent was specially designed and manufactured for the small diameter coronary vessels. A computational and experimental method of research on biomechanics of vascular stent was presented. Computational simulation of stent deployment expanded by the balloon based on nonlinear finite element analysis was performed including large displacement and deformation, geometric and material nonlinearity. The experimental platform of in vitro stent expansion based on the machine vision technology was established and the image processing software was developed. The fabricated stent was tested on the assembled experimental equipments. Matching between the computational and experimental results was quite satisfactory. The experimental scheme provides powerful support for the computational analysis of stent biomechanics.

This paper expatiated on the process of numerical computing of brain electric field in electroencephalogram (EEG). Based on boundary element method (BEM), a 3D reconstruction of computing model is presented first, which is the premise of BEM computing in EEG. A simple but efficient triangular mesh generation method with constrained points is developed, furthermore, a mesh subdivision method is also put forward. Forward computation of EEG is investigated and acceptable results are obtained with the simulated experiments by using these methods.

In this paper, we present a novel multi-stage algorithm for 3D medical image segmentation that is inspired by an improved Fast Marching method and a morphological reconstruction algorithm. The segmentation procedure consists of three steps: Connectivity Reduction, Hybrid segmentation, and Region recovery. The approach is tested on CT cardiac and MRI brain images, to demonstrate the effectiveness and accuracy of the technique. In order to validate this segmentation algorithm, a novel Radial Distance Based Validation (RDBV) method is proposed that provides a global accuracy (GA) measure. GA is calculated based on Local Radial Distance Errors (LRDE), where measured errors are along radii emitted from points along the skeleton of the object rather than the centroid, in order to accommodate more complicated organ structures. Using this GA measure, our results demonstrate that this multi-stage segmentation is fast and accurate, achieving approximately the same segmentation result as the watershed method, but with a processing speed of 3-5 times faster.

In this paper, a new approach on image registration is presented. We introduce a novel conception- normal vector information (NVI) – to evaluate the similarity between two images. NVI method takes advantage of the relationship between voxels in the image to extract the normal vector (NV) information of each voxel. Firstly, NVI criterion is presented. Then, based on the criterion, we find that NVI related metric has a quite perfect global optimal value on transformation parameter ranges. Finally, registration examples which are based on NVI criterion are provided. The result implies that the feature of smooth value distribution and one global optimal value that NVI metric has makes the optimization procedure much easier to be implemented in image registration.

Global exponential stability of non-autonomous delayed neural networks is discussed. A new sufficient condition ensuring the exponential stability for this type of neural networks by constructing a new suitable Lyapunov functionals. Since the condition does not impose boundedness on activation functions, it is less restrictive than some established in the earlier literature.

This paper introduces a prediction method for time series that is based on the multi-resolution analysis of wavelets (MRA). The MRA is better able to decompose the non-stationary time series of nonlinear systems into different components, allowing a better separation of the general trend terms, the periodic terms and the random fluctuation terms. By applying the most suitable prediction methods(for example, the neural networks method, cosine approximation, or the ARMA model) to the components under different resolutions, this new prediction method produces more accurate prediction results. The new approach is then applied to a real example – the BRENT oil price time series – to demonstrates its usefulness and validity.

Multi-layer perceptrons (MLPs) have been widely used in classification and regression task. How to improve the training speed of MLPs has been an interesting field of research. Instead of the classical method, we try to train MLPs by a MiniMin model which can ensure that the weights of the last layer are optimal at each step. Significant improvement on training speed has been made using our method for several big benchmark data sets.

Independent component analysis (ICA) has been applied in many fields of signal processing and many ICA learning algorithms have been proposed from different perspectives. However, there is still a lack of a deep mathematical theory to describe the ICA learning algorithm or problem, especially in the cases of both super- and sub-Gaussian sources. In this paper, from the point of view of the one-bit-matching principle, we propose two adaptive matching learning algorithms for the general ICA problem. It is shown by the simulation experiments that the adaptive matching learning algorithms can efficiently solve the ICA problem with both super- and sub-Gaussian sources and outperform the typical existing ICA algorithms in certain aspects.

Using genetic programming (GP) integrated with nonlinear parameter estimation we can identify the model for avermectin process. In order to reduce the effect caused by bloating which appears when a GP run stagnates in the later period, a fitness function with a double penalty strategy is proposed. GP with this penalty strategy is less sensitive to the choice of penalty parameters and compromises the fitness and the complexity of an individual, so the method can save considerable amounts of computational effort and find models with better quality. In addition, we combine the mechanism knowledge of the fermentation in GP to increase the quality of population and the convergence speed. Experiments prove that this method outperforms standard GP in reducing computational effort and finding better models more quickly.

Computational methods detecting the transcription factor binding sites (TFBS) remain one of the most intriguing and challenging subjects in bioinformatics. Gibbs sampling is essentially a heuristic method, and it is easy to trap into a nonoptimal “local maximum”. To overcome this problem and to improve the accuracy and sensitivity of the algorithm, we present an improved Gibbs sampling strategy MPWMGMS to search for TFBS. We have tested MPWMGMS and other existing Gibbs sampling algorithms on simulated data and real biological data sets with regulatory elements. The results indicate that MPWMGMS has better performance than other methods to a great extent in accuracy and sensitivity of finding true TFBS.

In this paper, a novel Fisher criterion is introduced and shown to be equivalent to the traditional Fisher criterion. Based on this new Fisher criterion and simultaneous diagonalization technique, a

S

t

-subspace Fisher discriminant (

S

t

-SFD) method is developed to deal with the small sample size (S3) problem in face recognition. The proposed method overcomes some drawbacks of existing LDA based algorithms. Also, our method has good computational complexity. Two public available databases, namely ORL and FERET databases, are exploited to evaluate the proposed algorithm. Comparing with existing LDA-based methods in solving the S3 problem, the proposed

S

t

-SFD method gives the best performance.

In this paper, an emotion recognition system for mandarin speech is presented. Five basic human emotions including angry, fear, happy, neutral and sad are investigated. The recognizer is based on neural network with OCON and ACON architecture. Some novel feature selection methods are also added as optional tool to enhance the efficiency and classification accuracy. The system can train speaker dependent emotion speech model through online emotional utterance recording. Experiment results show that emotion can be recognized through neural network model, the best mean accuracy is 86.7%. In addition, the feature selection module is effective to reduce the compute load and increase the generalization ability of the recognizer.

In this paper, we propose ubiFloorII, a novel floor-based user identification system to recognize humans based on their walking pattern such as stride length, dynamic range, foot angle, and stance and swing time. To obtain users walking pattern from their gait, we deployed photo interrupter sensors instead of switch sensors used in ubiFloorI. We developed a software module to extract walking pattern from users’ gait. For user identification, we employed neural network trained with users’ walking samples. We achieved about 96% recognition accuracy using this floor-based approach. The ubiFloorII system may be used to automatically and transparently identify users in home-like environments.

When a radial-basis function neural network (RBFNN) is used for pattern classification, the problem involves designing the topology of RBFNN and also its centers and widths. In this paper, we present a particle swarm optimization (PSO) learning algorithm to automate the design of RBF networks, to solve pattern classification problems. Simulation results for benchmark problems in the pattern classification area show that the PSO-RBF outperforms two other learning algorithms in terms of network size and generalization performance.

Patchouli is a common used traditional Chinese herbal medicine for flatulence or vomit with a long history in China and some other Asian countries. Patchoulis of different geographical origins usually have different curative effects. In order to evaluate their qualities, it’s very necessary to discriminate them. Our objective of this study is to develop a nondestructive and accurate identification method. The results in this paper showed that it’s difficult to use conventional infrared spectra and second derivative spectra directly to distinguish them. But it’s quite easy to use two-dimensional infrared (2D IR) correlation spectra, especially after applying wavelet transform process, to discriminate them. The resolution of the 2D IR spectrum is improved obviously after wavelet transform process, more peaks appear and the peaks become quite clear and separate. The differences of the 2D IR spectra become rather remarkable. In this way, it’s not difficult to discriminate the patchoulis of different geographical origins. This will be a nondestructive, economical and rapid way to distinguish complicated mixture like traditional Chinese medicines. Combined 2D IR and wavelet transform, Fourier transform infrared spectroscopy would become more powerful in analysis and discrimination.

This paper presents a new approach for human identification at a distance using gait recognition. Binarized silhouette of a motion object is represented by 1-D signals which are the basic image features called the distance vectors. The distance vectors are differences between the bounding box and silhouette, and extracted using four view directions to silhouette. Based on normalized correlation on the distance vectors, gait cycle estimation is first performed to extract the gait cycle. Second, eigenspace transformation based on PCA is applied to time-varying distance vectors and then Mahalanobis and normalized Euclidean distances based supervised pattern classification is finally performed in the lower-dimensional eigenspace for human identification. Experimental results on two main database demonstrate that the right person in top two matches 100% of the times for the cases where training and testing sets corresponds to the walking styles for data set of 25 people, and other data set of 22 people.

This paper presents a framework of HMM parameter adaptation technique for improving automatic speech recognition (ASR) performance in the noisy environments, which online combines the clean hidden Markov models (HMMs) with the noise model. Based on the given composite HMM corresponding to the initial recognition pass result and truncated vector Taylor series, the noise model in the cepstral domain is updated and refined using iterative Expectation-Maximization (EM) algorithm under maximum likelihood (ML) criterion. Experiments results show that the presented approach in this paper is found to greatly improve recognition performance under mismatched conditions.

This paper presents a method that automatically recognizes the shoe’s outsole products into model type, which flows through the conveyor belts from right to left. The interlaced pixels are displayed when we use the NTSC based camera in experiments. So, we require a suitable post-processing. For the purpose of this processing, it decides to find rectangle region of object by thresholding after removing interlaced pixels using de-interlacing method. And then, after rectangle region is separated into blocks through edge detection, we calculates pixel number per each block, re-classifies using its average, and classifies products into model type.

An R-functions based classification approach along with a regularization framework is proposed. The abnormal software process detection problem was used as the test bed. The R-functions based classification method is termed as the R-cloud method. The approach was validated both on synthetic and real-world data. Regularization allows to achieve good generalization and classification performance. In addition, the R-cloud approach gives the benefit of the analytical representation of the decision boundary. The introductory study on practical use of the R-cloud classifiers yielded promising results. The prototyping has shown that application of the R-functions based pattern recognition technique is a significant and practical tool for fault detection in providing fault tolerant computing.

According to drawbacks of available algorithms,a new hierarchical and multiresolution approach to the detection and classification of scene breaks in video sequences is presented in this paper. This method gives a unified framework for different shots by using the multi-resolution analysis as mainstay. Firstly, the video clips are cut by FCM clustering method, then the fade and dissolve are detected using ECR and SCD algorithm in the high components and the low components respectively by integer-to-integer wavelet transform. Finally, according to the cut and the former detection of gradual changes, we use motion vectors in high-components of 3D-WT to detect the wipe transition. Experimental results with real video clips about 7

h

demonstrate that our method can detect and classify a variety of scene breaks, including cuts, fades and dissolves, even in sequences involving significant motions and flash.

Locality preserving projections (LPP) can find an embedding that preserves local information and discriminates data well. However, only one projection matrix over the whole data is not enough to discriminate complex data. In this paper, we proposed locality preserving projections mixture models (LPP mixtures), where the set of all data were partitioned into several clusters and a projection matrix for each cluster was obtained. In each cluster, We performed LPP via QR-decomposition, which is efficient computationally in under-sampled situations. Its theoretical foundation was presented. Experiments on a synthetic data set and the Yale face database showed the superiority of LPP mixtures.

In this paper, a new line-based camera calibration technique is presented. Firstly, the relation between dimensional line parameters and projective line parameters is deduced. Using more than three dimensional lines which are not parallel with each other, the camera intrinsic parameters can be determined by solving linear equations. Experiment results show that in same noise level the mean attained by this method is more closer to factual value and MSE is smaller by about 65% than classic point-based technique.

Video shot boundary detection (SBD) is an important step in many video applications. In this paper, previous temporal multi-resolution analysis (TMRA) framework was extended by first using SVM (Supported Vector Machines) classify the video frames within a sliding window into normal frames, gradual transition frames and CUT frames, then clustering the classified frames into different shot categories. The experimental result on ground truth, which has about 26 hours (13,344 shots) news video clips, shows that the new framework has relatively good accuracy for the detection of shot boundaries. It basically solves the difficulties of shot boundaries detection caused by sub-window technique in video. The framework also greatly improves the accuracy of gradual transitions.

In this paper, a Devanagari script recognition scheme based on a novel algorithm is proposed. Devanagari script poses new challenges in the field of pattern recognition primarily due to the highly cursive nature of the script seen across its diverse character set. In the proposed algorithm, the character is initially subjected to a simple noise removal filter. Based on a reference co-ordinate system, the significant contours of the character are extracted and characterized as a contour set. The recognition of the character involves comparing these contour sets with those in the enrolled database. The matching of these contour sets is achieved by characterizing each contour based on its length, its relative position in the reference co-ordinate system and an interpolation scheme which eliminates displacement errors. In the Devanagari script, similar contour sets are observed among few characters, hence this method helps to filter out disparate characters and narrow down the possibilities to a limited set. The next step involves focusing on the subtle yet vital differences between the similar contours in this limited set. This is done by a prioritization scheme which concentrates only on those portions of character which reflect its uniqueness. The major challenge in developing the proposed scheme lay in striking the right balance between definiteness and flexibility to derive optimal solutions for out of sample data. Experimental results show the validity and efficiency of the developed scheme for recognition of characters of this script.

The paper introduces the method of probabilistic neural network (PNN) and its classifying principle. It constructs two PNN structures which are used to recognize both the two patterns and the three patterns respectively. The structure of the two patterns classification of PNN is used to classify the 106 listed companies of China in 2000 into two groups. The classification accuracy rate is 87.74%. The structure of the three patterns classification of PNN is used to classify the 96 listed companies of China in 2000 into three groups. The classification accuracy rate is 85.42%.

Fingerprint recognition and verification are always the key issues in intelligent technology and information security. Extraction of fingerprint ridge lines is a critical pre-processing step in fingerprint identification applications. Although existing algorithms for fingerprint extraction work well on good-quality images. Their performance decrease when handling poor-quality images. This paper addresses the ridge line extraction problem as curve tracking processes under the framework of probabilistic tracking. Each ridge line is modeled as sequential frames of a continuous curve and then traced by standard CONDENSATION algorithm in the area of computer vision. Additionally, local directional image is rectified with a feedback technique after each tracking step to improve the accuracy. The experimental results are compared with those obtained through existing well-known algorithms, such as local-binarization and sampling-tracing methods. In spite of greater computational complexity, the method proposed performs better both in terms of efficiency and robustness.

Human gait recognition is the process of identifying individuals by their walking manners. The gait as one of newly coming biometrics has recently gained more and more interests from computer vision researchers. In this paper, we propose a new method for model-free recognition of gait based on silhouette in computer vision sequences. The silhouette shape is represented by a novel approach which includes not only the spatial body contour but also the temporal information. First, a background subtraction is used to separate objects from background. Then, we represent the spatial shape of walker and their motion by the temporal matrix, and use Discrete Fourier analysis to analyze the gait feature. The nearest neighbor classifier is used to distinguish the different gaits of human. The performance of our approach is tested using different gait databases. Recognition results show this approach is efficient.

This paper presents a microcalcification patterns recognition method based autoassociator and classifier to detect the breast cancer. It studies the autoassociative and classification abilities of a neural network approach to classify the microcalcification patterns into Benign and Malignant using some certain image structure features. The proposed technique used the combination of two kinds of neural networks, autoassociator and classifier to analyze the microcalcification. It could obtain 88% classification rate for testing dataset and 100% classification rate for training dataset.

This paper presents an improved method which is suitable for gradient-threshold edge detectors. The proposed method takes into account the basic characteristics of the human visual system (HVS) and precisely determines the local masking regions for the edges with arbitrary shape according to the image content. Then the gradient image is masked with the luminance and the activity of local image before edge labelling. The experimental results show that the edge images obtained by our algorithm are more consistent with the perceptive edge images.

A novel technique for shape coding of a digital object based on its inner and/or outer isothetic polygonal shape, is proposed. The method uses a multigrid background for several applications including digital image visualization and retrieval with varying levels of accuracy. The elegance of the proposed shape code lies in capturing the shape of the objects present in an image from their gross appearances to finer details with a set of hierarchical isothetic polygons that tightly envelop (cover) the objects from outside (inside). Experimental results demonstrate the strength and efficiency of the proposed scheme.

This paper addresses how intelligent characters, having learning capability based on the neural network technology, automatically adapt to environmental changes in computer games. Our adaptation solution includes an autonomous adaptation scheme and a cooperative adaptation scheme. With the autonomous adaptation scheme, each intelligent character steadily assesses changes of its game environment while taking into consideration recently earned scores, and initiates a new learning process when a change is detected. Intelligent characters may confront various opponents in many computer games. When each intelligent character has fought with just part of the opponents, the cooperative adaptation scheme, based on a genetic algorithm, creates new intelligent characters by composing their partial knowledge of the existing intelligent characters. The experimental results show that intelligent characters can properly accommodate to the changes with the proposed schemes.

One of the greatest challenges for dependable service-oriented software systems of next generation is coping with the complexity of required adaptation or reaction to the detected unforeseen vulnerability attacks. To this end, autonomic system[1] has been advocated as a way to design self-protective systems to defend against malicious attacks or cascading failures. However, other initiatives such as the self-regenerative system[2] adopt the biological-inspired [2, 3]notions such as natural diversity and self-immune as a main strategy to achieve the robust and adaptable self-protection. Based on an ongoing research into self-regenerative programming model, this paper presents a knowledge-centric approach for supporting the runtime automated generation of software adapters for cross-standard service activation; and argues the importance of application of a semantic knowledge to extract the notion of self-regenerative adaptation from the previous polyarchical middleware implementation. The benefit of this will be the production of a customizable self-regenerative adaptation service; and also, support for abstraction integration between domain of similar interests or others in a high-level management directed towards building autonomic systems in a large domain of interest.

In this study, we will employ a multi-agent for searching and extraction of data. We will use Integrator Agent based on CORBA architecture for the proposed model on hierarchical Clustering and Association Rule Discovery (HCARD). The model will address the inadequacy of other data mining tools in processing performance and efficiency when use for knowledge discovery. The result revealed faster searching using the agents. Our experiment also shows that the HCARD generated isolated but imperative association rules which in return could be practically explained for decision making purposes. Shorter processing time had been noted in computing for smaller clusters implying ideal processing period than dealing with the entire dataset.

Filled function method is a type of efficient methods to solve global optimization problems arisen in non-convex programming. In this paper, a new class of filled functions is proposed. This class of filled functions has only one adjustable parameter

a

. Several examples of this class of filled functions with specified parameter values are given, which contain the filled functions proposed in [3] and [4]. These examples show this class of filled functions contains more simple functions, therefore this class of filled functions have better computability. An algorithm employing the proposed filled function is presented, and numerical experiments show that the proposed filled functions are efficient.

Both a concept of the optimal set of elementary siphons and a deadlock prevention policy based on integer programming are presented to solve deadlock problems arising in flexible manufacturing systems(FMS). Furthermore, an algorithm based on modified particle swarm optimization(PSO) is illustrated to show its efficiency to deal with such problems. Numerical simulation shows that this policy can minimize the number of newly additional control places and arcs while improving the dynamic performance of the resultant system.

For real-time computer-controlled systems, a control task does not have a fixed period but a range of periods in which control performance varies. Hence for multiple tasks scheduled on a single processor, to consider the optimization design of sampling periods in the co-design of control and scheduling is necessary to improve the control performance and use limited computing resource efficiently. In this paper, the mathematic description of the optimization problem of designing periods is presented, and the optimization solution using evolution strategy is proposed. The performances of proposed solution are revealed via simulation studies. Simulation shows that the optimization design of sampling periods can be implemented by using the evolution strategy method.

A multi-objective evolutionary algorithm (MOEA) based approach to Power System Stabilizer (PSS) and Static Var Compensators (SVC) tuning has been investigated in this paper. The coordinated design problem of PSS and SVC is formulated as a multi-objective optimization problem, in which the system response is optimized by minimizing several system-behavior measure criterions. MOEA is employed to search optimal controller parameters.Design of the multi-objective optimization aims to find out the Pareto optimal solution which is a set of possible optimal solutions for controller parameters. And effectiveness of the proposed control scheme has been demonstrated in a multiple power system.