New Challenges in Applied Intelligence Technologies

In this comparative study the concept of common ground is introduced for understanding the process of knowledge sharing by human and robot, and applied to “questioning and answering” task during object identification experiments. Various cases of human-human and human-robot interactions were performed and the interaction patterns among these groups were investigated. In comparison to Human-human interaction, the performance of robot was determined by both its expertise level as well as its human partner’s expertise level. This suggests that the robot’s intelligence alone is not sufficient to communicate with human users.

In this paper, we make a first step towards a formal model of dialogue and argumentation for a multi-agent (model-based) diagnostic system. We shall discuss some of the issues in multi-agent cooperative fault diagnosis, the theories of communicating agents and their reasoning capabilities. We propose a Partial Information State (PIS)-based framework for dialogue and argumentation. We shall employ a three-valued based nonmonotonic logic for representing and reasoning about partial information. We show via an example that the system can easily be customized to handle distributed problem-solving tasks.

Learning in negotiation is fundamental for understanding human behaviors as well as for developing new solution concepts. Elsewhere, negotiation behaviors, in which the characters such as Conciliatory (Con), Neutral (Neu), or Aggressive (Agg) define a ‘psychological’ aspect of the negotiator personality, play an important role. In this paper, first, a brief description of SISINE (Integrated System of Simulation for Negotiation) project, which aims to develop innovative teaching methodology of negotiation skills, is given. Second, a negotiation approach essentially based on the escalation level and negotiator personality is suggested for SISINE. In fact, the escalation level defines gradually different negotiation stages from agreement to interruption. Afterwards, negotiation behaviors acquired by reinforcement Q-learning and Neural Networks (NN) under supervised learning are developed. Then, behavior results which display the suggested approach ability to provide negotiators with a first intelligence level are presented. Finally, a discussion is given to evaluate this first intelligence level.

An

Artificial Life

environment -

getALife

- is proposed, whose major aim is to provide a framework to evaluate single and multi-agent systems and evolutionary approaches to the development of reinforcement learning algorithms. The environment is based on a predator-prey scenario, with multiple species and where individuals are mainly characterized by their decision modules and genetic information. The platform is quite powerful, flexible, modular, visually attractive, easy to program and to use, making an interesting tool both to research and teaching. Two applications based on

getALife

are provided: the evaluation of a

Neural Network

based decision module with evolutionary learning and the development of a children’s game.

This paper proposes a new cooperation business model in which hybrid traders exist. We define hybrid traders as new traders on the Internet. Hybrid traders can become both buyers and sellers. We assume that hybrid traders do not have enough money. To buy items cheaply, hybrid traders cooperate with other traders. In regard to buying items, we consider a volume discount-based trading. We propose a mechanism in which trader cooperates, buys in a lot of goods, and increases own utility. Our mechanism adopts side payment to promote increasingly cooperation with traders. Cooperative traders commit participation based on a value of side payment. We extend mechanism which hybrid traders deal with multiple items. This mechanism shows new decision of side payment and proposer’s strategy.

Decision making in a multi-agent environment has continued to be one of the most formidable challenges for AI researchers. Increasing the efficiency of predictors is an essential task, especially in a substrate like robosoccer where a misclassification can cost dearly. It is also necessary for the agent to perform well, irrespective of the nature of testing data, generated in a markovian fashion. For this reason, we apply AdaBoost (Adaptive Boosting) algorithm onto support vector machines which helps us achieve a better generalization performance than normal Support Vector Machines (SVM) and a better efficiency when compared to other adaboosted neural networks. To illustrate the concept, we propose a highly efficient decision predictor for low-level behavior in robosoccer using Adaboosted SVM (AdSVM). Through experiments, we have proved that the proposed agent model has outwitted existing neural networks and SVM in classifying two-class data of any nature in a multi-agent environment like robosoccer.

We use closed pattern mining to discover user preferences in appointments in order to build structured solutions for a calendar assistant. Our choice of closed patterns as a user preference representation is based on both theoretical and practical considerations supported by Formal Concept Analysis. We simulated interaction with a calendar application using 16 months of real data from a user’s calendar to evaluate the accuracy and consistency of suggestions, in order to determine the best data mining and solution generation techniques from a range of available methods. The best performing data mining method was then compared with decision tree learning, the best machine learning algorithm in this domain. The results show that our data mining method based on closed patterns converges faster than decision tree learning, whilst generating only consistent solutions. Thus closed pattern mining is a better technique for generating appointment attributes in the calendar domain.

On the Web, Electronic Commerce is widely thriving with development of the Web technology. However, users still have trouble finding products that will find their desires. In recent researches, they have introduced many kinds of method for Recommender systems, but these system still have problems which are based on concrete attributes of the products and a complex users model. Within this paper, we introduce a new technique, Goal Oriented Recommendation, which works even when users do not want exactly products that they are looking for. Moreover, the system processes users’ input (e.g. “I’m going to have dinner with my boss” or “I’m looking for my wife’s birthday presents”) with a own concept dictionary which contains a occasion word and a person word. The system can recommend items based on users’ desire, if users input their desire.

As electronic commerce thrives, products on the Web are increasing and consumers are having trouble finding products that meet their desires. Many recommender systems help users find desired products, but not all use communication among users positively for recommendations. We propose a recommender system with a web-chat interface to make an environment where users can talk with each other. By observing conversation in a web-chat interface in real-time and always recognizing user interests, the system can recommend products based on conversation contents.

This paper illustrated an Interface Agent which works as an assistant between the users and FAQ systems to retrieve FAQs on the domain of Personal Computer. It can effectively tackle the problems associated with traditional FAQ retrieval systems. Specifically, we addressed how ontology helps interface agents to provide better FAQ services and related algorithms described in details. Our preliminary experimentation demonstrated that user intention and focus of up to eighty percent of the user queries can be correctly understood by the system, and accordingly provided the query solutions with higher user satisfaction.

This paper discusses subsystem responsible for providing human resource adaptation through software-supported training in an agent-based virtual organization. Attention is focused on the requirements, functionalities and components of this subsystem and its interactions with other parts of the system.

A main trend in CP is to define a standard modeling language. This challenge is not a minor matter whose success may depend on many years of experimental steps. Several concerns must be studied such as the simplicity, the level of expressiveness and a suitable solver-independent architecture. In this paper we introduce the

s-COMMA

modeling language and its execution platform. In this approach a constraint language including extension mechanisms has been carefully fused with object-oriented capabilities in order to provide a considerable level of expressiveness and simplicity. The system is supported by a solver-independent three-layered architecture where models can be mapped to many solvers. We believe the work done on

s-COMMA

represents a concrete step on the pursuit of a standard constraint modeling language.

Architectural solutions to information integration have extensively appeared during the last years, mostly from the federated system research field. Some of these solutions were created to deal with geographic information, whose inherent features make the integration process particularly complex. Among others, the use of ontologies has been proposed as a way of supporting an automated integration. However, how to specify and use a geographic ontology is not so clear in this context. In this paper, we introduce an ontology-based architectural solution as an extension of a federated system (Information Broker) built by the Italian Agency for Environmental Protection and Technical Services (APAT). Our extension is aimed at improving integration by adding semantic features through the use of ontologies and the ISO 19100 standards.

We generalize the QSQR evaluation method to give a set-oriented depth-first evaluation method for Horn knowledge bases. The resulting procedure closely simulates SLD-resolution (to take advantages of the goal-directed approach) and highly exploits set-at-a-time tabling. Our generalized QSQR evaluation procedure is sound, complete, and tight. It does not use adornments and annotations. To deal with function symbols, our procedure uses iterative deepening search which iteratively increases term depth bound for atoms occurring in the computation. When the term depth bound is fixed, our evaluation procedure runs in polynomial time in the size of extensional relations.

This paper provides an analysis of cues to identify Arabic vowels. A new algorithm for vowel identification has been developed that uses formant frequencies. The algorithm extracts the formants of already segmented recitation audio files and recognizes the vowels on the basis of these extracted formants. The investigation has been done in context of recitation principles of Holy Quran which are commonly known as Tajweed rules. Primary objective of this work is to be able to identify zabar /a/, zair /e/ and pesh /u/ mistakes of the recitor during the recitation. Acoustic Analysis was performed on 150 samples of different recitors and a corpus comprising recitation of five experts was used to validate the results. The vowel identification system developed here has shown up to 90% average accuracy on continuous speech files comprising around 1000 vowels.

Nowadays, Artificial Neural Networks are being widely used in the representation of physical processes. Once trained, the nets are capable to solve unprecedented situations, keeping tolerable errors in their outputs. However, humans cannot assimilate the knowledge kept by these networks, since such knowledge is implicitly represented by their structure and connection weights. Recently, the FCANN method, based in Formal Concept Analysis, has been proposed as a new approach in order to extract, represent and understand the behavior of the process through rules. In this work, it is presented an adaptation of the FCANN method to extract more comprehensible variables relationships, obtaining a reduced and more interesting set of rules related to a predefined domain parameters subset, which provides a better analysis of the knowledge extracted from the neural networks without the necessity of a posteriori implications mining. As case study the approach FCANN will be applied in solar energy system.

In this paper, an original scheme for video similarity detection is proposed in order to establish correspondence between two video sequences. This scheme consists first to summarize the visual contents of a video sequence in a small set of images. Each image is then modeled, by an Attributed Relational Graph (ARG), as the composition of salient objects with specific spatial relationship. Matching two video sequences is thereby reduced to the ARG similarity problem. The proposed approach offers a principled way to define the ARG similarity that accounts for both the attribute and topological differences of the two considered ARGs. Indeed, we proposed herein a cost-efficient solution to find the best alignment between two ARGs. This consists to the minimization of a similarity measure between the two graphs using dynamic programming. This measure can be considered as a matching rate which can be very useful for Content Based Video Retrieval (CBVR) applications. The suggested scheme was preliminary tested on real-world databases and very promising results were observed.

Even though the possession of a high CMM level undoubtedly implies prestige and competitive advantages for a software development organisation, its attainment may imply a considerable economic burden because of potentially necessary audits. It is therefore very interesting to minimise the costs by paying only for the truly indispensable audits. This article proposes a Knowledge-Based System that makes it possible to evaluate an organisation at a determined CMM level and as such limit the services of an auditor to those cases in which the system’s response complies with the requested CMM level and the necessary associated skills. This clearly implies an important cost reduction for audits with a negative result. The design of this system is based on the CommonKADS methodology, and its implementation was carried out with the Clips tool.

This work proposes a novel approach to Soft Sensor modelling, where the Soft Sensor is built by a set of experts which are artificial neural networks with randomly generated topology. For each of the experts a meta neural network is trained, the gating Artificial Neural Network. The role of the gating network is to learn the performance of the experts in dependency on the input data samples. The final prediction of the Soft Sensor is a weighted sum of the individual experts predictions. The proposed meta-learning method is evaluated on two different process industry data sets.

This paper addresses a classification method on users opinions for a content provider to grasp useful opinions from large amount of answers in open-ended questionnaires. Users opinions are categorized into known and unknown opinions by identifying known ones with typical words pattern-based extraction technique, however, the unknown opinions database still includes known opinions. Therefore our proposed method here introduces the category-based dictionary generated from the known opinions database and judges an opinion from two criteria; category typical words involvement and category words coincidence. We also discuss applied results of this method to mobile game content users opinions.

Preparing a database to train classifiers for identification of musical instruments in audio files is very important, especially in a case of sounds of the same pitch, when a dominating instrument is most difficult to identify. Since it is infeasible to prepare a data set representing all possible ever recorded mixes, we had to reduce the number of sounds in our research to a reasonable size. In this paper, our data set represents sounds of selected instruments of the same octave, with additions of artificial sounds of broadband spectra for training, and additions of sounds of other instruments for testing purposes. We tested various levels of added sounds taking into consideration only equal steps in logarithmic scale which are more suitable for amplitude comparison than linear one. Additionally, since musical instruments can be classified hierarchically, experiments for groups of instruments representing particular nodes of such hierarchy have been also performed. The set-up of training and testing sets, as well as experiments on classification of the instrument dominating in the sound file, are presented and discussed in this paper.

Algorithm selection problem is a common problem when we solve optimization problems. To cope with it we have proposed a hybrid system of metaheuristics that intelligently combines different strategies using a coordinator based on Fuzzy Logic. In this paper we study the impact of Fuzzy Logic in the behaviour of this hybrid system. In order to do that we perform some test to study the impact of an important parameter, the

α

− 

cut

used in the fuzzy engine of the system, demonstrating how the variations on this parameter may change the performance of the system with different kind of instances.

Due to the importance of solving discrete combinatorial problems in so many fields, constraint programming was developed. One of the most challenging concerns is the representation of these problems through the well known Constraint Satisfaction Problem (CSP) framework. In this paper we present a tool that facilitates the specification of constraint applications via CSP. Based on the Object Constraint Language (OCL), the tool provides a CSP template that can be easily specialized to describe a wide range of constraint problems. More precisely, we enhance OCL with new keywords to be able to express constraints and requirements of CSP. From the CSP specification, our tool automatically generates the constraint and solution graphs. Afterwards, we demonstrate the usability and efficiency of our tool on large size problems.

In this paper, we propose a method of discovering the best combination between helpers and elders under much restriction based on multi-agent systems. Scheduling problem is defined as task assignment problem. Computational costs increase exponentially according to the number of restrictions. We present a new allocation algorithm which uses agent based experienced technique to detect an efficient allocation. The agent technology is one field of the artificial intelligence. It is promising in the negotiation among users who have various roles because of the problem solving. Our proposing method is based on constraint satisfaction problem. We also derive the optimal solution with reducing the solution space. To use our proposed calculation method, elderly person can serve an effective service allocation since business managers can improve their tasks efficiently. We show an example that we found efficient combination from a situation in which five elders and five helpers are given several parameters. By using our proposed method, a business manager for home-care can deliver appropriate services in both city and rural town.

Cancer screening is a method of preventing cancer by early detecting and treating abnormalities. One of the most critical screening phase is invitation planning since screening resources are limited and there are many people to invite. For this reason, smart resource allocation approaches are needed.

In the paper, we propose and compare two solutions for smart invitation plan definition, one based on greedy approaches and one based on Constraint Programming techniques that enable the definition of the optimal invitation plan.

This paper presents a novel mathematical model of a closed-loop layout problem with unequal-sized facilities. This problem belongs to a class of combinatorial optimization and NP-hard problems. Obtaining an optimal solution for this complex, large-sized problem in reasonable computational time by using traditional approaches and is extremely difficult. Therefore, we propose three well-known meta-heuristics, namely genetic algorithm (GA), ant colony optimization (ACO), and simulated annealing (SA), to solve the closed-loop layout problem. These algorithms report near-optimal and promising solutions in a short period of time because of their efficiency. The computational results obtained by these algorithms are compared with the results reported by the Lingo 8.0 software package. Finally among our three proposed meta-heuristics, the output of SA is better than other two algorithms and the Lingo.

In recent years we have witnessed a tremendous growth of communication networks resulted in a large variety of combinatorial optimization problems. One of these problems is the terminal assignment problem. In this paper, we propose a genetic algorithm employing multiple crossover and mutation operators for solving the well-known terminal assignment problem. Two sets of available crossover and mutation operators are established initially. In each generation a crossover method is selected for recombination and a mutation method is selected for mutation based on the amount fitness improvements achieved over a number of previous operations (recombinations/mutations). We use tournament selection for this purpose. Simulation results with the different methods implemented are compared.

This paper presents a new mathematical model for an expected maximal covering location problem (EMCLP) that optimizes both location and allocation decisions. In real-world cases, traveling or lead times may change over the period of time. It is assumed that the traveling time between customers and emergency centers has an exponential distribution function. If this uncertain time is less than a critical time, the customer can be allocated to that emergency center (facility), which yields more flexibility for results and the proposed model. In this paper, we present a stochastic nonlinear model that selects the best location of emergency centers, and allocation decisions maximizing the total expected demand covered. To solve such a difficult model, an efficient hybrid method based on the simulation and genetic algorithm is proposed. Finally, some numerical examples are illustrated to show the effectiveness of this proposed method.

Elective Subject is one of important issues as education program in University. Students can declare their preferences directly by selecting it. In most of university, to allocate elective subjects to the students, university staffs poll students the lectures they want to take. However, due to the limitation of time and number of staffs, the hearing investigation includes the reason and the intention in which students select the lectures. Some students sometimes take a lecture for their career, for academical interest, and for assimilation of knowledge. However, some students might take the lecture following the crowd and take the lecture as Mickey Mouse. The latter case is undesirable for the higher education. To solve the problem, in this paper, we propose a new multi-step lecture allocation method based on students preferences and university intentions. Our protocol consists of the three steps of negotiations and three types of allocations. (1) The university warns the students who have never take a certain compulsory subject yet. The students can choose whether they attend the lecture or not. If the students answer they attend the lecture, the students are allocated to the lecture by priority. (2) The students inform the university of their reasons to take the lecture. The university allocates the lectures to the students based on their reasons. (3) They negotiate about the exchange of lectures to increase students’ utilities with each student. Our protocol realizes the high performance of allocation compared with brute force algorithm and reducing computational costs compared with optimizations.

The alignment and comparison of DNA, RNA and Protein sequences is one of the most common and important tasks in Bioinformatics. However, due to the size and complexity of the search space involved, the search for the best possible alignment for a set of sequences is not trivial. Genetic Algorithms have a predisposition for optimizing general combinatorial problems and therefore are serious candidates for solving multiple sequence alignment tasks. We have designed a Genetic Algorithm for this purpose: AlineaGA. We have tested AlineaGA with representative sequence sets of the hemoglobin family. We also present the achieved results so as the comparisons performed with results provided by T-COFFEE.

Support Vector Machines constitute a Machine Learning technique originally designed for the solution of two-class problems. This paper investigates and proposes strategies for the generalization of SVMs to problems with more than two classes. The focus of this work is on strategies that decompose the original multiclass problem into binary subtasks, whose outputs are combined. The proposed strategies aim to investigate the adaptation of the decompositions for each multiclass application considered, using information of the performance obtained in its solution or extracted from its examples. The implemented algorithms were evaluated using benchmark datasets and real applications from the Bioinformatics domain. Among the benefits observed is the obtainment of simpler decompositions, which require less binary classifiers in the multiclass solution.

This paper presents specification of the agent monitoring system, responsible for supervision of the network elements (switches, workstations). Mechanism described below, handles passive role in aspects of the network administration. It is dedicated for monitoring predefined system statistics of the network members (hosts, switches, routers, etc.), like CPU usage, memory usage, network traffic.

Intrusion detection systems (IDSs) are exposed to highly dynamic and demanding environments. Moreover, analysis engines based on either signature recognition or anomaly detection most often are large modules which consume a significant amount of resources. In distributed IDSs individual monitoring entities are capable of detecting local intrusions based on performed observations. In this paper we propose that agents are assigned some collection of monitoring tasks, which need varying amount of computational resources. These needs vary over time causing occasional overload of single monitoring entities and resulting in a need for tasks’ reassignment. An analysis performed on distant objects brings an additional load over the system. Therefore to reduce an additional network traffic invoked by multi-object reassignment a method of single object’s delegation is proposed, based on neighbours’ lists and distances between monitoring agents and observed objects. Designed solution has been implemented with JADE and compared with a random delegation solution.

An intelligent service strategy is one of the key elements in ensuring quality of service in a computer network. This paper presents a new analytical model for investigating a linked computer network with blocking and feedback (service according to the HOL priority scheme.) It describes behaviour of a computer network exposed to an open Markovian queuing model with blocking. The model illustrated below is very accurate, derived directly from a two-dimensional state graph and a set of steady-state equations, followed by calculations of Quality of Service (QoS) parameters. Collected numerical results indicate that the proposed open queuing network model with blocking and feedback can provide accurate performance estimates of a network. In our examples, the performance is calculated and numerically illustrated by varying buffer capacities, regulating intensity of the input flow and altering feedback probability. In addition, blocking probabilities in such network are calculated.

Developing models for fault localization in HDL designs has been an active research area in recent years. Whereas research on circuit verification is typically conducted on Verilog programs, research on fault localization has recently focused on the VHDL domain. The research presented herein focuses on fault localization models for Verilog designs and thus promotes the investigation of the relationships between models for property verification and fault localization. Primarily we focus on two novel contributions. First, this article points out notable semantic differences between VHDL and Verilog models and discusses its implications for fault localizations. Secondly, we advance existing work by incorporating multiple testcases and provide first empirical results obtained from the the ISCAS 89 benchmarks indicating our novel technique’s applicability for real world designs.

We investigate the hypothesis that it is possible to detect from the End User License Agreement (EULA) if the associated software hosts spyware. We apply 15 learning algorithms on a data set consisting of 100 applications with classified EULAs. The results show that 13 algorithms are significantly more accurate than random guessing. Thus, we conclude that the hypothesis can be accepted. Based on the results, we present a novel tool that can be used to prevent spyware by automatically halting application installers and classifying the EULA, giving users the opportunity to make an informed choice about whether to continue with the installation. We discuss positive and negative aspects of this prevention approach and suggest a method for evaluating candidate algorithms for a future implementation.

We consider the problem of estimating the clustering of nodes in wireless sensor networks (WSNs). A solution to this problem is proposed, which uses Least Squares Support Vector Machines (LS-SVM). Using mixtures of kernels and the image energy distribution of the sensor field surface, we have been solved the clustering problem in WSNs. Some computer experiments for the simulated sensor fields are carried out. Through comparing with classical clustering scheme we state that LS-SVM method has a better improvement in clustering accuracy in these networks.