Skip to main content
main-content

Über dieses Buch

The two-volume set LNAI 8467 and LNAI 8468 constitutes the refereed proceedings of the 13th International Conference on Artificial Intelligence and Soft Computing, ICAISC 2014, held in Zakopane, Poland in June 2014. The 139 revised full papers presented in the volumes, were carefully reviewed and selected from 331 submissions. The 69 papers included in the first volume are focused on the following topical sections: Neural Networks and Their Applications, Fuzzy Systems and Their Applications, Evolutionary Algorithms and Their Applications, Classification and Estimation, Computer Vision, Image and Speech Analysis and Special Session 3: Intelligent Methods in Databases. The 71 papers in the second volume are organized in the following subjects: Data Mining, Bioinformatics, Biometrics and Medical Applications, Agent Systems, Robotics and Control, Artificial Intelligence in Modeling and Simulation, Various Problems of Artificial Intelligence, Special Session 2: Machine Learning for Visual Information Analysis and Security, Special Session 1: Applications and Properties of Fuzzy Reasoning and Calculus and Clustering.

Inhaltsverzeichnis

Frontmatter

Data Mining

Frontmatter

Visual Dictionary Pruning Using Mutual Information and Information Gain

Feature selection methods are often applied to many machine learning problems, one of the applications involves selecting most informative Visual Words for image categorization task. In Bag of Visual Words framework, image is represented as vector of frequencies of Visual Words, typically of length from hundreds to thousands elements. A dictionary of Visual Words is produced from image keypoints detected by SIFT algorithm and quantized into words by k-means clustering. In the paper we use Mutual Information and Information Gain as methods for selecting these words that are the most important for efficient image classification. There are four novel methods, which expand use of classic Mutual Information and Information Gain in line with our previous feature selection methods. We consider two basic selection strategies: one-vs-all and one-vs-one, as well as multi class and multi attribute value problems. The experimental session we have conducted has shown a positive effect of our modification, when applied to image classification by Support Vector Machines. The results showed that visual word selection based on modified Mutual Information in most cases wins over methods based on Information Gain.

Piotr Artiemjew, Przemysław Górecki

Mining Telecommunication Networks to Enhance Customer Lifetime Predictions

Customer retention has become a necessity in many markets, including mobile telecommunications. As it becomes easier for customers to switch providers, the providers seek to improve prediction models in an effort to intervene with potential churners. Many studies have evaluated different models seeking any improvement to prediction accuracy. This study proposes that the attributes, not the model, need to be reconsidered. By representing call detail records as a social network of customers, network attributes can be extracted for use in various traditional prediction models. The use of network attributes exhibits a significant increase in the area under the receiver operating curve (AUC) when compared to using just individual customer attributes.

Aimée Backiel, Bart Baesens, Gerda Claeskens

A Note on Machine Learning Approach to Analyze the Results of Pairwise Comparison Based Parametric Evaluation of Research Units

This paper presents an attempt at an analysis of parametric evaluation of research units with machine learning toolkit. The main goal was to investigate if the rules of evaluation can be expressed in a readable, transparent, and easy to interpret way. A further attempt was made at investigating consistency of the applied procedure and presentation of some observed anomalies.

Mateusz Baran, Konrad Kułakowski, Antoni Ligęza

Bagging of Instance Selection Algorithms

The paper presents bagging ensembles of instance selection algorithms. We use bagging to improve instance selection. The improvement comprises data compression and prediction accuracy. The examined instance selection algorithms for classification are ENN, CNN, RNG and GE and for regression are the developed by us Generalized CNN and Generalized ENN algorithms. Results of the comparative experimental study performed using different configurations on several datasets shows that the approachbased on bagging allowed for significant improvement, especially in terms of data compression.

Marcin Blachnik, Mirosław Kordos

Essential Attributes Generation for Some Data Mining Tasks

In this paper, we introduce a new approach referred to as Essential Attributes Generation (EAG) to reduce the dimensionality of multidimensional real-valued data series. We form a new representation of the original data. The approach is based on the concept of essential attributes generated by a multilayer neural network. The EAG generates a vector of real valued new attributes which form the compressed representation of the original data. The attributes are synthetic, and while not being directly interpretable, they still retain important features of the original data series. The approach has found applications to classification as well as clustering tasks.

Maciej Krawczak, Grażyna Szkatuła

Visualizing Random Forest with Self-Organising Map

Random Forest (RF) is a powerful ensemble method for classification and regression tasks. It consists of decision trees set. Although, a single tree is well interpretable for human, the ensemble of trees is a black-box model. The popular technique to look inside the RF model is to visualize a RF proximity matrix obtained on data samples with Multidimensional Scaling (MDS) method. Herein, we present a novel method based on Self-Organising Maps (SOM) for revealing intrinsic relationships in data that lay inside the RF used for classification tasks. We propose an algorithm to learn the SOM with the proximity matrix obtained from the RF. The visualization of RF proximity matrix with MDS and SOM is compared. What is more, the SOM learned with the RF proximity matrix has better classification accuracy in comparison to SOM learned with Euclidean distance. Presented approach enables better understanding of the RF and additionally improves accuracy of the SOM.

Piotr Płoński, Krzysztof Zaremba

B-Spline Smoothing of Feature Vectors in Nonnegative Matrix Factorization

Nonnegative Matrix Factorization (NMF) captures nonnegative, sparse and parts-based feature vectors from the set of observed nonnegative vectors. In many applications, the features are also expected to be locally smooth. To incorporate the information on the local smoothness to the optimization process, we assume that the features vectors are conical combinations of higher degree B-splines with a given number of knots. Due to this approach the computational complexity of the optimization process does not increase considerably with respect to the standard NMF model. The numerical experiments, which were carried out for the blind spectral unmixing problem, demonstrate the robustness of the proposed method.

Rafał Zdunek, Andrzej Cichocki, Tatsuya Yokota

Variants and Performances of Novel Direct Learning Algorithms for L2 Support Vector Machines

The paper introduces a novel Direct L2 Support Vector Machine (DL2 SVM) classifier and presents the performances of its 4 variants on 12 different binary and multiclass datasets. Direct L2 SVM avoids solving quadratic programming (QP) problem and it solves the Nonnegative Least Squares (NNLS) task instead, which, unlike the related iterative algorithms, produces an impeccably accurate results. Solutions obtained by NNLS and QP are equal but NNLS needs much less CPU time. The comprehensive DL2 SVM model, as well as its three variants, are devised. The similarities with, and differences in respect to, LS SVM and proximal SVMs are pointed at too. The four DL2 SVM models performances are compared in terms of accuracy, percentage of support vectors and CPU time. A strict nested cross-validation (double resampling) is used in all experiments.

Ljiljana Zigic, Vojislav Kecman

Bioinformatics, Biometrics and Medical Applications

Frontmatter

Evolving Parameters for a Noisy Biological System – The Impact of Alternative Approaches

In this contribution we seek to evolve viable parameter values for a small-scale biological

network motif

concerned with bacterial nutrient uptake and metabolism. We use two different evolutionary approaches with the model: implicit and explicit. Our results reveal that significantly different characteristics of both efficiency and timescale emerge in the resulting evolved systems depending on the which particular approach is used.

David J. Barnes, Dominique Chu

Classification of EEG Signals Using Vector Quantization

Proper identification and classification of the EEG data still pauses a problem in the field of brain diagnosis. However, the application of such algorithm is almost unlimited as they may be involved in applications such as, brain computer interface for controlling of prosthesis, wheelchair, etc.. In this paper we are focusing on applying data compression in the classification of EEG signals. We combine a vector quantization and the normalized compression distance for proper classification of a finger movement data.

Petr Berek, Michal Prilepok, Jan Platos, Vaclav Snasel

Offline Text-Independent Handwriting Identification and Shape Modeling via Probabilistic Nodes Combination

Proposed method, called Probabilistic Nodes Combination (PNC), is the method of 2D curve modeling and handwriting identification by using the set of key points. Nodes are treated as characteristic points of signature or handwriting for modeling and writer recognition. Identification of handwritten letters or symbols need modeling and the model of each individual symbol or character is built by a choice of probability distribution function and nodes combination. PNC modeling via nodes combination and parameter

γ

as probability distribution function enables curve parameterization and interpolation for each specific letter or symbol. Two-dimensional curve is modeled and interpolated via nodes combination and different functions as continuous probability distribution functions: polynomial, sine, cosine, tangent, cotangent, logarithm, exponent, arc sin, arc cos, arc tan, arc cot or power function.

Dariusz Jacek Jakóbczak

Computer-Aided System for Automatic Classification of Suspicious Lesions in Breast Ultrasound Images

In this research, a new method for automatic detection of suspected breast cancer lesions using ultrasound images is proposed. In this fully automated method, the best de-noising technique from among several considered is selected, a new segmentation based on fuzzy logic is proposed and detection of lesions based on morphological features and texture features is considered. We also consider correlation among ultrasound images taken from different angles and use it to improve detection.

Behnam Karimi, Adam Krzyżak

The Classifier for Prediction of Peri-operative Complications in Cervical Cancer Treatment

This paper addresses the problem of creating a new classifier as highly interpretable fuzzy rule-based system, based on the analytical theory of fuzzy modeling and gene expression programming. This approach is applied to solve the prediction problem of peri-operative complications of radical hysterectomy in patients with cervical cancer. The developed classifier has the form of the set of fuzzy metarules, which are readable for the medical community, and additionally, is accurate enough. The consequents of the metarules describe the presence or absence of peri-operative complications. For the construction of the classifier we can use the fuzzified, binarized or both types of the attributes. We also compare the efficiency of our model with the decision trees and C5 algorithm.

Jacek Kluska, Maciej Kusy, Bogdan Obrzut

Face Classification Based on Linguistic Description of Facial Features

This paper presents an artificial intelligence approach towards classification of persons based on verbal descriptions of their facial features. Frame knowledge representation, fuzzy sets, fuzzy IF-THEN rules, and fuzzy granulation are employed. Features of face elements (nose, eyes, etc.) are extracted by use of existing detection techniques, such as measurements of horizontal and vertical sizes. Linguistic variables that correspond to fuzzy sets, representing selected facial features, are applied in the frames and fuzzy rules. Linguistic values defined by the fuzzy sets conform the terminology applied by law enforcement to create an eyewitness verbal description. Classification results are illustrated in three cases of the system’s input: facial composites (sketches) created by an artist, images (digital pictures) from a face database, and verbal descriptions.

Damian Kurach, Danuta Rutkowska, Elisabeth Rakus-Andersson

Impact of Bayesian Network Model Structure on the Accuracy of Medical Diagnostic Systems

While Bayesian network models may contain a handful of numerical parameters that are important for their quality, several empirical studies have confirmed that overall precision of their probabilities is not crucial. In this paper, we study the impact of the structure of a Bayesian network on the precision of medical diagnostic systems. We show that also the structure is not that important – diagnostic accuracy of several medical diagnostic models changes minimally when we subject their structures to such transformations as arc removal and arc reversal.

Agnieszka Oniśko, Marek J. Druzdzel

A New Three-Dimensional Facial Landmarks in Recognition

In recent years, the number of biometric solutions based on 3D face images has increased rapidly. Such solutions provide a much more accurate alternative to those using flat images; however, they are much more complex. In this paper, we present subsequent results of our research on a new representation of characteristic points for the 3D face. As a comparative method the standard PCA is applied.

Sebastian Pabiasz, Janusz T. Starczewski, Antonino Marvuglia

Computer-Aided Off-Line Diagnosis of Epileptic Seizures

EEG signal analysis is commonly used by skilled neurologists as a useful tool in the diagnosis of specific neurological dysfunction. To greatly facilitate the diagnostic process and improve the efficiency of decision making decision support systems are developed based on expert knowledge. This paper presents the design of a computer system supporting seizure detection, based on real EEG records. The system is based on modern signal processing tools that allow for time-frequency representation of the analyzed signal. The proposed solution should be treated as a decision support computer system. The system has been designed to facilitate the rapid detection of characteristic graphoelements to effectively detect epileptic seizures. The proposed solution can have a significant impact on an accuracy and speed in the analysis of EEG signals, which may significantly shorten the time of making diagnosis trials. The proposed system is based on studies using real EEG records of patients with epilepsy as well as healthy subjects prepared in collaboration with the medical staff of the Ward of Neurology and Strokes of the Provincial Hospital of Zielona Góra, Poland.

Grzegorz Rutkowski, Krzysztof Patan

Active Region Approach for Segmentation of Medical Images

Active region models are methods for automatic image segmentation. In this paper, the method is examined using various medical images (heart, brain and liver). The quality measure, taken for evaluation of the method is based on combination of two measures used for classifiers. The energy of the region is based on statistical features of initial region.

Pawel Tracz, Piotr S. Szczepaniak

SCM-driven Tree View for Microarray Data

Eisen’s tree view is a useful tool for clustering and displaying of microarray gene expression data. In Eisen’s tree view system, a hierarchical method is used for clustering data. However, some useful information in gene expression data may not be well drawn when a hierarchical clustering is directly used in Eisen’s tree view. In this paper, we embed the similarity-based clustering method (SCM) into the tree view system so that microarray data can be re-organized according to the structure of data. The created SCM-driven tree view can give a better dendrogram display for microarray gene expression data with more useful information.

Hsun-Chih Kuo, Miin-Shen Yang, Jenn-Hwai Yang, Yen-Chi Chen

New Method for Dynamic Signature Verification Using Hybrid Partitioning

Dynamic signature is behavioural biometric attribute which is commonly used to identity verification. Methods based on the partitioning are one of the types of methods for identity verification using signature biometric attribute. These methods divide trajectories of the signature into parts and during verification phase compare created fragments of trajectories in each partition. Partitioning is performed on the basis of values of signals describing dynamics of signing process (e.g. pen velocity or pen pressure). In this paper we propose a new method for dynamic signature verification using hybrid partitioning. Partitions in the proposed method can be interpreted as, for example, high velocity in the first phase of the signing process or low pressure in the final phase of the signing process. Our method assumes use of all partitions during classification process and our classifier is based on the flexible neuro-fuzzy system of the Mamdani type. Simulations were performed using public SVC2004 dynamic signature database.

Marcin Zalasiński, Krzysztof Cpałka, Meng Joo Er

New Method for Dynamic Signature Verification Based on Global Features

Identity verification based on the dynamic signatures is commonly known issue of biometrics. This process is usually done using methods belonging to one of three approaches: global approach, local function based approach and regional function based approach. In this paper we focus on global features based approach which uses the so called global features extracted from the signatures. We present a new method of global features selection, which are used in the training and classification phase in a context of an individual. Proposed method bases on the evolutionary algorithm. Moreover, in the classification phase we propose a flexible neuro-fuzzy classifier of the Mamdani type. Our method was tested using the SVC2004 public on-line signature database.

Marcin Zalasiński, Krzysztof Cpałka, Yoichi Hayashi

Agent Systems, Robotics and Control

Frontmatter

Investigating the Rate of Failure of Asynchronous Gathering in a Swarm of Fat Robots with Limited Visibility

In the present paper we investigate the failure probability of the asynchronous gathering task in a mobile robot swarm that contains weak robots (oblivious, with limited visibility, without global navigation and communication). We are modelling fat robots, which are represented as solid discs. We performed numerous computer simulations in order to measure the rate of failure of gathering using extended Ando’s gathering algorithm [1]. The physical parameters of the simulations are based on our previous experiments on image processing based kin recognition method using Surveyor SRV-1 robots [11] [12]. It was obtained that the computational time and the travelling speed of the robots affect very strongly the rate of gathering success. If we apply SRV-1 robots with the referred kin recognition method and highest possible travelling speed, then the rate of failure is very close to 100 percent. While, reducing the travelling speed with a factor of 1/20 (or increasing the computational performance to 20 times the original) results in much better success rates of gathering. Besides, we have found that the failure rate increases together with the number of robots.

Kálmán Bolla, Tamás Kovács, Gábor Fazekas

Modeling Context-Aware and Agent-Ready Systems for the Outdoor Smart Lighting

Smart lighting systems considered as context-awareness systems are challenging. The use of advanced technology for street lighting allows to achieve a number of potential benefits of improving the efficiency of lighting, enhance the ability to monitor and control street lighting. A context-aware based system in architecture for street lighting control dealing with intelligent software applications in pervasive computing has been proposed. Some diagrams in the UML language are extended by some elements of the CML language in order to provide possibilities to design and verify behaviour of context-aware-based systems.

Radosław Klimek, Grzegorz Rogus

Problem of Agents Cooperation in Heterogeneous Graph-Based Knowledge Environment

The important aspect of performance of a multi-agent system deployed in highly non-homogeneous environment is ensuring data confidentiality. In such an environment each agent resides in its local ecosystem characterized by its own data structure and semantics. Moreover agents may also extend their knowledge by sending appropriate queries to other agents. Thus a knowledge diffuses through the system so it may cause unauthorized data accesses. In this article we propose the run-time method of discovering such data leaks and the method of preventing such events.

Leszek Kotulski, Adam Sędziwy, Barbara Strug

Managing Machine’s Motivations

This paper presents concepts for the development and management of motivations in learning agents, which are critical for motivated learning. We suggest that an agent must be equipped with a mechanism referred to as a

nonspecific formative process

to trigger higher level motivations. Resource and action related motivations are discussed as examples of implementing such process in a virtual world learning scenario.

Janusz Starzyk, James Graham, Leszek Puzio

Globalised Dual Heuristic Dynamic Programming in Tracking Control of the Wheeled Mobile Robot

The paper presents an application of the Approximate Dynamic Programming algorithm in Globalised Dual Heuristic Dynamic Programming configuration in the tracking control problem of the wheeled mobile robot Pioneer 2-DX. The Globalised Dual Heuristic Dynamic Programming algorithm is realised in the form of two structures, the actor and the critic, that can be implemented in the form of any adaptive algorithm, e.g. Artificial Neural Networks. The actor generates the suboptimal control law, the critic approximates the value function and its difference with respect to the states, what is equal to evaluation of the realised control law. The discrete tracking control system is composed of the Globalised Dual Heuristic Dynamic Programming algorithm, the PD controller and the supervisory term, which structure derives from the stability analysis realised using the Lapunov stability theorem. The proposed control system works on-line and its performance was verified using the wheeled mobile robot Pioneer 2-DX.

Marcin Szuster

Fuzzy Sensor-Based Navigation with Neural Tracking Control of the Wheeled Mobile Robot

Navigation of the wheeled mobile robot in the unknown environment with simultaneous realisation of the generated trajectory, is one of the most challenging and up to date problems in the modern mobile robotics. In the article a new approach is presented to a collision-free trajectory generating for a wheeled mobile robot, realised in a form of the hierarchical control system with two layers. The first layer is a tracking control system, where the Neuro-Dynamic Programming algorithm in the Dual Heuristic Dynamic Programming configuration was applied. The second layer is a trajectory generator where the Fuzzy Logic systems were used. The presented control system generates and realises trajectory of the wheeled mobile robot within the complex task of goal-seeking and obstacle avoiding. The proposed hierarchical control system works on-line, its performance was verified using the wheeled mobile robot Pioneer 2-DX.

Marcin Szuster, Zenon Hendzel, Andrzej Burghardt

Artificial Intelligence in Modeling and Simulation

Frontmatter

Fuzzy and Neural Rotor Resistance Estimator for Vector Controlled Induction Motor Drives

This paper contributes to improving the dynamic performance of indirect vector controlled induction motor drives. This command requires the rotor resistance; the variation of this parameter could distort the decoupling between the flux and torque and, consequently, lead to deterioration of performance. To overcome this problem two intelligent approaches have been introduced to estimate the rotor resistance namely fuzzy logic and artificial neural networks. These estimators process the information from the rotational speed, the stator currents and voltages. The performances of the two intelligent approaches are investigated and compared in simulation. The results show that the neural rotor resistance estimator is reliable and highly effective in the resistance identification relative to fuzzy rotor resistance estimator of induction motor drives.

Moulay Rachid Douiri, Ouissam Belghazi, Mohamed Cherkaoui

ALMM Solver - A Tool for Optimization Problems

The aim of our paper is to present the concept and structure of a software tool named the ALMM Solver. The goal of the solver is to generate solutions for discrete optimization problems, in particular for NP-hard problems. The solver is based on Algebraic Logical Meta-Model of Multistage Decision Process (ALMM of MDP) methodology, which is briefly described in the paper. Functionality and modular structure of the ALMM Solver is presented. SimOpt, the core module of the solver, is described in detail. Some possible future advances regarding the solver are also given.

Ewa Dudek-Dyduch, Edyta Kucharska, Lidia Dutkiewicz, Krzysztof Rączka

Tournament Searching Method for Optimization of the Forecasting Model Based on the Nadaraya-Watson Estimator

In the article the tournament searching method is used for optimization of the forecasting model based on the Nadaraya-Watson estimator. This is a nonparametric regression model useful for forecasting the nonstationary in mean and variance time series with multiple seasonal cycles and trend. The tournament searching is a stochastic global optimization algorithm which is easy to use and competitive to other stochastic methods such as evolutionary algorithms. Three types of tournament searching algorithms are proposed: for estimation of the forecasting model parameters (continuous optimization), for the predictor selection (binary optimization) and for both predictor selection and parameter estimation (mixed binary-continuous optimization). The effectiveness of the proposed approach is illustrated through applications to electrical load forecasting and compared with other optimization methods: grid search method, genetic and evolutionary algorithms, and sequential methods of feature selection. Application examples confirm good properties of tournament searching.

Grzegorz Dudek

A New Algorithm for Identification of Significant Operating Points Using Swarm Intelligence

The paper presents a novel algorithm for identification of significant operating points from non-invasive identification of nonlinear dynamic objects. In the proposed algorithm to identify the unknown parameters of nonlinear dynamic objects in different significant operating points, swarm intelligence supported by a genetic algorithm is used for optimization in continuous domain. Moreover, we propose a new weighted approximation error measure which eliminates the problem of the measurements obtained from non-significant areas. This measure significantly accelerates the process of the parameters identification in comparison with the same algorithm without weights. Performed simulations prove efficiency of the novel algorithm.

Piotr Dziwiński, Łukasz Bartczuk, Andrzej Przybył, Eduard D. Avedyan

Artificial Bee Colony Algorithm Used for Reconstructing the Heat Flux Density in the Solidification Process

Scope of the paper is the procedure reconstructing the heat flux density in the solidification on the grounds of temperature measurements in selected points of the cast. Elaborated method is based on two procedures: finite difference method with application of the generalized alternating phase truncation method used for solving the appropriate direct solidification problem and the Artificial Bee Colony algorithm used for minimizing some functional representing the crucial part of the procedure.

Edyta Hetmaniok, Damian Słota, Adam Zielonka

Simulations of Credibility Evaluation and Learning in a Web 2.0 Community

Since the emergence of Web 2.0, the idea of online knowledge sharing has been gaining attention of researchers and online communities. We can observe the popularity of such services on Wikipedia and numerous Q&A systems, in which ordinary users can explicitly ask questions and provide answers thus raise their expertise level by learning from others. Users dynamically switch between roles of content producer and content consumer. This paper applies game-theoretic approach to study how different community member profiles and reputation can affect the learning process and, in consequence, credibility of the provided information.

Grzegorz Kowalik, Paulina Adamska, Radosław Nielek, Adam Wierzbicki

Optimization of Composite Structures Using Bio-inspired Methods

The paper deals with an application of the artificial immune system (AIS) and the particle swarm optimizer (PSO) to the optimization problems. The AIS and PSO are applied to optimize of stacking sequence of plies in composites. The optimization task is formulated as maximization of minimal difference between the first five eigenfrequencies and the external excitation frequency. Recently, immune and swarm methods have found various applications in mechanics, and also in structural optimization. The AIS is a computational adaptive system inspired by the principles, processes and mechanisms of biological immune systems. The algorithms typically use the characteristics of the immune systems like learning and memory to simulate and solve a problem in a computational manner. The swarm algorithms are based on the models of the animals social behaviours: moving and living in the groups. The main advantage of the AIS and PSO, contrary to gradient methods of optimization, is the fact that they do not need any information about the gradient of fitness function. The numerical examples demonstrate that the new method based on immune and particle computation is an effective technique for solving computer aided optimal design.

Arkadiusz Poteralski, Mirosław Szczepanik, Witold Beluch, Tadeusz Burczyński

Applying Metamodels and Sequential Sampling for Constrained Optimization of Process Operations

This paper presents a framework for nonlinear constrained optimization of complex systems, in which the objective function and the constraints are represented by black box functions. The proposed approach replaces the complex nonlinear model based on first principles with Kriging metamodels. Coupled to Kriging, the “Constrained Expected Improvement” technique and a sequential sampling strategy are used to explore the metamodels, in order to find global solutions for the constrained nonlinear optimization problem. The methodology has been tested and compared with classical optimization procedures based on sequential quadratic programming. Both have been applied to three mathematical examples, and to a case study of chemical process operation optimization. The proposed framework shows accurate solutions and significant reduction in the computational time.

Ahmed Shokry, Antonio Espuna

Various Problems of Artificial Intelligence

Frontmatter

A New Measure of Conflict and Hybrid Combination Rules in the Evidence Theory

Based on the critical analysis of methods for evaluation of conflict between basic probability assignments (

bpa

s) to be combined and combination rules in the Dempster-Shafer theory of evidence, a new simple, but reliable method for the evaluation of conflict between combining

bpa

s is proposed and analysed. Using some critical examples, it is shown that the proposed approach performs better than Dempster’s rule and the known hybrid rule based on the weighted sum of conjunction and disjunction operators. It is shown that in the case of small conflict, the use of averaging rule for combination of

bpa

s seems to be a best choice.

Ludmila Dymova, Pavel Sevastjanov, Kamil Tkacz, Tatyana Cheherava

On Measuring Association between Groups of Rankings in Recommender Systems

A measure of association between two groups of rankings is proposed. The suggested measure possesses some interesting properties which make it useful in recommender systems and some other possible applications. In particular, it aggregates the bipolar information taking into account both the strength of the correlation and its sign. Simultaneously, applied in collaborative filtering, it rewards strong association which is a desired property in making meaningful recommendations to a user.

Hanna Łącka, Przemysław Grzegorzewski

An Arduino-Simulink-Control System for Modern Hand Protheses

Despite the significant amount of research on the application of myoelectric research for upper limb prostheses control [1] and advances in signal processing and classification methods for myoeletric signals (MES), patient satisfaction and acceptance for modern hand prostheses is lacking [2]. This is partly due to missing intuitive and natural control possibilities for accessing the various grip patterns that are available with current prostheses models on the market. As a step towards easy prototyping and seamless integration of a wide variety of prostheses, we present a system based on the Arduino microcontroller platform. With adaptable Simulink

TM

models and a wide number of libraries for the Arduino IDE, the system allows electromyographic (EMG) processing as well as basic classification for actuating both basic hand models and more advanced hand prostheses. Complex classifier models can be trained with a PC-based MATLAB

TM

application prior to microcontroller operation.

Andreas Attenberger, Klaus Buchenrieder

Solving Timetabling Problems on GPU

This paper concerns the application of a parallel

tabu search

algorithm to solve the general problem of

timetabling

. The problem of timetabling (also known as scheduling) was first expressed as a graph coloring problem and then good approximate solutions were obtained with use of concurrent metaheuristic algorithm for GPU (

Graphics Processing Unit

).

Wojciech Bożejko, Łukasz Gniewkowski, Mieczysław Wodecki

Scheduling Problem with Uncertain Parameters in Just in Time System

This paper tackles a stochastic version of single machine scheduling with random processing times and due dates. The objective function is to find a schedule of jobs which minimizes the weighted sum of earliness and tardiness penalties, taking into consideration probabilistic modelling of the tasks processing times. Since the problem is NP-hard, it is not possible to find optimal values for large scales of the problem in a reasonable run time, therefore, a tabu search meta-heuristic is provided. Computational experiments demonstrate that the proposed tabu search algorithm is strongly capable of finding near optimal stability solutions with a very low gap.

Wojciech Bożejko, Paweł Rajba, Mieczysław Wodecki

Variable Neighborhood Search for Non-deterministic Problems

A comparative analysis of several neighborhood structures is presented, including a variable neighborhood structure, which corresponds to a combination of the neighborhood structures evaluated in this paper. The performance of each neighborhood structure was tested using large random instances generated in this research and well-known benchmarks such as the Classical Symmetric Traveling Salesman Problem and the Unrelated Parallel Machines Problem. Experimental results show differences in the performance of the variable neighborhood search when it is applied to problems with differing complexity. Contrary to reports in literature about variable neighborhood searches, its performance varies according to the complexity of the problem.

Marco Antonio Cruz-Chávez, Alina Martínez-Oropeza, Jesús del Carmen Peralta-Abarca, Martín Heriberto Cruz-Rosales, Martín Martínez-Rangel

Emergent Phenomena in Constrained 3D Layout Design

While explored in the context of creativeness in computer aided design, emergence has hardly ever been mentioned as a phenomenon that may enhance optimization process in engineering design. This paper presents the original approach to constrained 3D component layout design problem that takes advantage of visual shape grammar computations, emergent phenomena and computational intelligence methods. Possible design solutions are generated with a use of a simple shape grammar. Design specific knowledge is represented in a form of goals and constraints and the search process is driven by an intelligent derivation controller. The presented framework is very general but in the same time flexible and easily adjustable to a specific problem domain.

Katarzyna Grzesiak-Kopeć, Maciej Ogorzałek

Brainy: A Machine Learning Library

Brainy is a newly created cross-platform machine learning library written in Java. It defines interfaces for common types of machine learning tasks and implementations of the most popular algorithms. Brainy utilizes a complex mathematical infrastructure which is also part of the library. The main difference compared to other ML libraries is the sophisticated system for feature definition and management. The design of the library is focused on efficiency, reliability, extensibility and simple usage. Brainy has been extensively used for research as well as commercial projects for major companies in Czech Republic and USA. Brainy is released under the GPL license and freely available from the project web page.

Michal Konkol

Optimisation of Character n-gram Profiles Method for Intrinsic Plagiarism Detection

The focus of the paper is to improve intrinsic plagiarism detection. The paper investigates and improves performance of character n-grams profiles method proposed by Stamatatos by tuning its parameter settings and proposing new modifications and rich feature sets. We raised the overall plagdet score from 24.67% to 33.41% for the PAN-PC09 corpus and from 18.83% to 26.66% for the PAN-PC11 corpus. Results are reported on PAN-PC09 and PAN-PC11 corpora, which are especially well suited for this task and were previously used in Plagiarism Analysis, Authorship Identification, and Near-Duplicate Detection (PAN) competitions.

Marcin Kuta, Jacek Kitowski

A Recommender System Based on Customer Reviews Mining

As e-commerce is becoming more and more popular, the number of different products reviews done by customer grows rapidly. The efficient method for automatic summarization of such reviews is required. The majority of existing approaches classify a review only whether the opinion is positive or negative. In the present paper we show how to extract product features from the set of the reviews to design feature based summaries of available opinions. These summaries, expressed in IF-set framework, are later used to recommend a customer the best product corresponding to his individual demands.

Paweł P. Ładyżyński, Przemysław Grzegorzewski

Complex System Analysis Using Softcomputing

The paper presents a usage of softcomputing methods to reliability and functional analysis of sophisticated complex systems. The approach is sensible since dependability parameters of the system are mostly approximated by experts instead of classical sources of data. The analyzed - Computer Information System

(CIS)

and Discrete Transport System

(DTS)

- are modelled using the unified structure - in functional sense. Results of numerical experiment performed on a test case scenario related to the reliability, economic and functional aspects using softcomputing are given. The presented approach allows reducing the problem of assumptions of reliability distributions and - this way - seems to be very interesting for real systems management and tuning.

Jacek Mazurkiewicz

Can the Generation of Test Cases for Unit Testing be Automated with Rules?

In this paper a proposal of a new black-box unit testing method based on decision tables is given. Its main part is an automatic generation of test cases using rule-based specification of a module. The tables containing rules are described in a formalized way using the XTT2 design method for rule-based systems. The paper also provides a presentation of a prototypical framework uses proposed method. This tool was designed as an Eclipse plugin which generates JUnit test cases. The proposed method can automate and improve the software testing process.

Grzegorz J. Nalepa, Krzysztof Kutt, Krzysztof Kaczor

Implementing a Supply Chain Management Policy System Based on Rough Set Theory

This paper presents a new method for raw material classification based on rough set theory. The classification method is used within the distribution network of a supply chain management system. An expert system is developed based a set of decision rules. The purpose of the expert system is to configure the distribution policies in order to reduce the transportation and storage costs as well as downtime risks.

Henryk Piech, Aleksandra Ptak, Ali Jannatpour

What Is the Primary Language?

This paper includes results of the research on the structure of the Primary Language of the human brain (as introduced by J. von Neumann in 1957). Two components have been investigated, Linguistic Geometry (LG) and the Algorithm of Discovery. I suggest that both components are mental realities “hard-wired” in the human brain. LG is a formal model of human reasoning about armed conflict, an evolutionary product of millions of years of human warfare. It was rediscovered via research on modeling and generalization of the human expert approach to playing chess and applied successfully to modern warfare. Experiences of development of LG have been instructive for discovering the Algorithm of Discovery, the foundation of all the discoveries throughout the history of humanity. This Algorithm is based on multiple thought experiments, which manifest themselves and are controlled by the mental visual streams.

Boris Stilman

Signal Randomness Measure for BSS Ensemble Predictors

In this article we present the application of novel noise measure in ensemble method based on blind signal separation methods. In this approach we decompose the set of models’ results into basis latent components with destructive or constructive impact on the prediction. The crucial step in such model aggregation is proper identification of destructive components which can be treated as noisy factors. Presented method assesses the randomness of signals using a new measure of variability which helps to compare analyzed signal with some typical noise models. The experiments performed on electric load data using different blind separation algorithms contributed to model improvements.

Ryszard Szupiluk, Tomasz Ząbkowski

An Incremental Map-Matching Algorithm Based on Hidden Markov Model

Map-matching algorithms aim at establishing a vehicle location on a road segment based on positioning data from a variety of sensors: GPS receivers, WiFi or cellular radios. They are integral part of various Intelligent Transportation Systems (ITS) including fleet management, vehicle tracking, navigation services, traffic monitoring and congestion detection. Our work was motivated by an idea of developing an algorithm that can be both utilized for tracking individual vehicles and for monitoring traffic in real-time. We propose a new incremental map-matching algorithm that constructs of a sequence of Hidden-Markov Models (HMMs). Starting from an initial HMM, the next models are developed by alternating operations: expansion and contraction. In the later, the map-matched trace is output. We discuss results of initial experiments conducted for 20 GPS traces, which to test algorithm robustness, were modified by introduction of noise and/or downsampled.

Piotr Szwed, Kamil Pekala

Machine Learning for Visual Information Analysis and Security

Frontmatter

Multi-class Classification: A Coding Based Space Partitioning

In this work we address the problem of multi-class classification in machine learning. In particular, we consider the coding approach which converts a multi-class problem to several binary classification problems by mapping the binary labeled space into several partitioned binary labeled spaces through binary channel codes. By modeling this learning problem as a communication channel, these codes are meant to have error correcting capabilities and thus performance improvement in classification. However, we argue that conventional coding schemes designed for communication systems do not treat the space partitioning problem optimally, because they are heedless of the partitioning behavior of underlying binary classifiers. We discuss an approach which is optimal in terms of space partitioning and advise it as a powerful tool towards multi-class classification. We then review the LDA, a known method for multi-class case and compare its performance with the proposed method. We run the experiments on synthetic data in several scenarios and then on a real database for face identification.

Sohrab Ferdowsi, Sviatoslav Voloshynovskiy, Marcin Gabryel, Marcin Korytkowski

From Single Image to List of Objects Based on Edge and Blob Detection

In this paper we present a new method for obtaining a list of interest objects from a single image. Our object extraction method works on two well known algorithms: the Canny edge detection method and the quadrilaterals detection. Our approach allows to select only the significant elements of the image. In addition, this method allows to filter out unnecessary key points in a simple way (for example obtained by the SIFT algorithm) from the background image. The effectiveness of the method is confirmed by experimental research.

Rafał Grycuk, Marcin Gabryel, Marcin Korytkowski, Rafał Scherer, Sviatoslav Voloshynovskiy

Robust Face Recognition by Group Sparse Representation That Uses Samples from List of Subjects

In this paper we consider group sparsity for robust face recognition. We propose a model for inducing group sparsity with no constraints on the definition of the structure of the group, coupled with locality constrained regularization. We formulate the problem as bounded distance regularized

L

2

norm minimization with group sparsity inducing, non-convex constrains. We apply convex relaxation and a branch and bound strategy to find an approximation to the original problem. The empirical results confirm that with this approach of deploying a very simple non-overlapping group structure we outperform several state-of-the-art sparse coding based image classification methods.

Dimche Kostadinov, Sviatoslav Voloshynovskiy, Sohrab Ferdowsi, Maurits Diephuis, Rafał Scherer

Using Facial Asymmetry Properties and Hidden Markov Models for Biometric Authentication in Security Systems

This work concerns the use of biometric features, resulting from the look of a face, for the authentication purposes. For this we propose several different methods of selection and feature analysis during face recognition. The description contains mainly the possibility of the analysis and in later stages also identity verification based on asymmetric facial features. The new authentication method has been introduced on the basis of designated characteristic points of face. The method includes propositions of our own algorithms of face detection, as well as face features extraction methods and their specific coding in the form of observation vectors and recognition using Hidden Markov Models.

Mariusz Kubanek, Dorota Smorawa, Mirosław Kurkowski

Spatial Keypoint Representation for Visual Object Retrieval

This paper presents a concept of an object pre-classification method based on image keypoints generated by the SURF algorithm. For this purpose, the method uses keypoints histograms for image serialization and next histograms tree representation to speed-up the comparison process. Presented method generates histograms for each image based on localization of generated keypoints. Each histogram contains 72 values computed from keypoints that correspond to sectors that slice the entire image. Sectors divide image in radial direction form center points of objects that are the subject of classification. Generated histograms allow to store information of the object shape and also allow to compare shapes efficiently by determining the deviation between histograms. Moreover, a tree structure generated from a set of image histograms allows to further speed up process of image comparison. In this approach each histogram is added to a tree as a branch. The sub tree is created in a reverse order. The last element of the lowest level stores the entire histogram. Each next upper element is a simplified version of its child. This approach allows to group histograms by their parent node and reduce the number of node comparisons. In case of not matched element, its entire subtree is omitted. The final result is a set of similar images that could be processed by more complex methods.

Tomasz Nowak, Patryk Najgebauer, Jakub Romanowski, Marcin Gabryel, Marcin Korytkowski, Rafał Scherer, Dimce Kostadinov

Applications and Properties of Fuzzy Reasoning and Calculus

Frontmatter

On Orientation Sensitive Defuzzification Functionals

The aim of the article is to investigate defuzzification functionals in the theory of Ordered Fuzzy Numbers (OFN). The model of OFN was introduced in 2002 to overcome drawbacks of classical (convex) fuzzy numbers. Each OFN is equipped with an additional feature – the orientation. New forms of defuzzification functionals are proposed which are sensitive to the orientation change.

Tomek Bednarek, Witold Kosiński, Katarzyna Węgrzyn-Wolska

The Linguistic Modeling of Fuzzy System as Multicriteria Evaluator for the Multicast Routing Algorithms

The paper presents the use of fuzzy system in multicriteria evaluation of algorithms that generate multicast trees and optimize realtime data transmission in computer networks. These algorithms take into account a number of factors such as: cost, bandwidth or delay, and their efficiency can be represented by total cost of multicast tree or average path’s cost in multicast tree [18]. However, there is a lack of accurate methods for comparing and evaluating these algorithms. In addition, it is difficult to identify with precision the weight of the criteria. The paper describes various proposals models underlying linguistic system that performs two-criteria assessment. These proposals show how to implement linguistic changes and their impact on the results of the fuzzy system.

Piotr Prokopowicz, Maciej Piechowiak, Piotr Kotlarz

Optimizing Inventory of a Firm under Fuzzy Data

The aim of the article is to propose some tools of inventory management. One is based on the so-called fixed order quantity model which takes into account several elements of inventory cost, such as ordering cost, transportation and storing costs, frozen capital cost, as well as extra discounts. The tool deals with fuzzy concepts represented by Ordered Fuzzy Numbers. The second tool takes into account the dynamics and works on the base of replenishment system. This tool can be regarded as a kind of controller.

Irena Sobol, Kurt Frischmuth, Dariusz Kacprzak, Witold Kosiński

An Approach to Cardinality of First Order Metasets

Metaset is a new approach to sets with partial membership relation. Metasets are designed to represent and process vague, imprecise data, similarly to fuzzy sets. They enable expressing fractional certainty of membership, equality, and other relations. Even though the general idea stems from and is firmly suited in the classical set theory, it is directed towards efficient computer implementations and applications.

In this paper we introduce the concept of cardinality for metasets and we investigate its basic properties. For simplicity we focus on the subclass of first order metasets however, the discussed ideas remain valid in general. We also present additional results obtained for finite first order metasets which are relevant for computer applications.

Bartłomiej Starosta

Fuzzy System for the Classification of Sounds of Birds Based on the Audio Descriptors

This paper presents an application of fuzzy systems for the classification of sounds coded by the selected MPEG-7 descriptors. The model of the fuzzy classification system is based on the audio descriptors for a few chosen species of birds: Great Spotted Woodpecker, Greylag, Goldfinch, Chaffinch. The paper proposes two fuzzy models that definitely differ by the description of the input linguistic variables. The results show, that both approaches are effective. However, second one is more flexible in a case of future expanding of the model with next descriptors or species of birds.

Krzysztof Tyburek, Piotr Prokopowicz, Piotr Kotlarz

Clustering

Frontmatter

Generalized Tree-Like Self-Organizing Neural Networks with Dynamically Defined Neighborhood for Cluster Analysis

The paper presents a generalization of self-organizing neural networks of spanning-tree-like structures and with dynamically defined neighborhood (SONNs with DDN, for short) for complex cluster-analysis problems. Our approach works in a fully-unsupervised way, i.e., it operates on unlabelled data and it does not require to predefine the number of clusters in a given data set. The generalized SONNs with DDN, in the course of learning, are able to disconnect their neuron structures into sub-structures and to reconnect some of them again as well as to adjust the overall number of neurons in the system. These features enable them to detect data clusters of virtually any shape and density including both volumetric ones and thin, shell-like ones. Moreover, the neurons in particular sub-networks create multi-point prototypes of the corresponding clusters. The operation of our approach has been tested using several diversified synthetic data sets and two benchmark data sets yielding very good results.

Marian B. Gorzałczany, Jakub Piekoszewski, Filip Rudziński

Nonnegative Matrix Factorization for Document Clustering: A Survey

Nonnegative Matrix Factorization (NMF) is a popular dimension reduction technique of clustering by extracting latent features from high-dimensional data and is widely used for text mining. Several optimization algorithms have been developed for NMF with different cost functions. In this paper we apply several methods of NMF that have been developed for data analysis. These methods vary in using different cost function for matrix factorization and different optimization algorithms for minimizing the cost function. Reuters Document Corpus is used for evaluating the performance of each method. The methods are compared with respect to their accuracy, entropy, purity and computational complexity and residual mean square root error. The most efficient methods in terms of each performance measure are also recognized.

Ehsan Hosseini-Asl, Jacek M. Zurada

Fuzzy c-Medoid Graph Clustering

We present a modified fuzzy c-medoid algorithm to find central objects in graphs. Initial cluster centres are determined by graph centrality measures. Cluster centres are fine-tuned by minimizing fuzzy-weighted geodesic distances calculated by Dijkstra’s algorithm. Cluster validity indices show significant improvement against fuzzy c-medoid clustering.

András Király, Ágnes Vathy-Fogarassy, János Abonyi

A Spectral Clustering Algorithm Based on Eigenvector Localization

This paper introduces the SpecLoc algorithm that performs clustering without pre-assigning the number of clusters. This is achieved by the use of a special property of matrix eigenvectors, called weak localization. The signless Laplacian matrix is created on the basis of a mutual neighbor graph. A new measure, introduced in this work, allows for selection of weakly localized eigenvectors. Experiments confirm good performance of the proposed algorithm for weakly separated groups of real datasets, including cancer gene expression matrices.

Małgorzata Lucińska

HiBi – The Algorithm of Biclustering the Discrete Data

The article presents the new algorithm for hierarchical biclustering:

HiBi

. It is dedicated to the analysis of the discrete data. The algorithm uses the set of exact biclusters as the input. In this approach results of exact biclustering algorithm

eBi

are used as the input. As a result of combining biclusters into the more general one,

HiBi

gives the set of inexact biclusters. The algorithm is hierarchical so the final result can be chosen after the algorithm performance. All experiments were performed on artificial datasets.

Marcin Michalak, Magdalena Lachor, Andrzej Polański

Asymmetric k-means Clustering of the Asymmetric Self-Organizing Map

In this paper, an asymmetric approach to clustering of the asymmetric Self-Organizing Map (SOM) is proposed. The clustering is performed using an improved asymmetric version of the well-known

k

-means algorithm. The improved asymmetric

k

-means algorithm is the second proposal of this paper. As a result, we obtain the two-stage fully-asymmetric data analysis technique. In this way, we maintain the structural consistency of the both utilized methods, because they are both formulated in asymmetric version, and consequently, they both properly adjust to asymmetric relationships in analyzed data. The results of our experiments confirm the effectiveness of the proposed approach.

Dominik Olszewski, Janusz Kacprzyk, Sławomir Zadrożny

DenClust: A Density Based Seed Selection Approach for K-Means

In this paper we present a clustering technique called DenClust that produces high quality initial seeds through a deterministic process without requiring an user input on the number of clusters

k

and the radius of the clusters

r

. The high quality seeds are given input to K-Means as the set of initial seeds to produce the final clusters. DenClust uses a density based approach for initial seed selection. It calculates the density of each record, where the density of a record is the number of records that have the minimum distances with the record. This approach is expected to produce high quality initial seeds for K-Means resulting in high quality clusters from a dataset. The performance of DenClust is compared with five (5) existing techniques namely CRUDAW, AGCUK, Simple K-means (SK), Basic Farthest Point Heuristic (BFPH) and New Farthest Point Heuristic (NFPH) in terms of three (3) external cluster evaluation criteria namely F-Measure, Entropy, Purity and two (2) internal cluster evaluation criteria namely Xie-Beni Index (XB) and Sum of Square Error (SSE). We use three (3) natural datasets that we obtain from the UCI machine learning repository. DenClust performs better than all five existing techniques in terms of all five evaluation criteria for all three datasets used in this study.

Md Anisur Rahman, Md Zahidul Islam, Terry Bossomaier

Exploiting Structural Information of Data in Active Learning

In recent years, the active learning algorithms have focused on combining correlation criterion and uncertainty criterion for evaluating instances. Although these criteria might be useful, applying these measures on whole input space globally may lead to inefficient selected instances for active learning. The proposed method takes advantage of clustering to partition input space to subspaces. Then it exploits both labeled and unlabeled data locally for selection of instances by using a graph-based active learning. We define a novel utility score for selecting clusters by combining uncertainty criterion, local entropy of clusters and the factor of contribution of each cluster in queries. Experimental results reveal an elevated performance as compared to several state of the art and widely used active learning strategies.

Maryam Shadloo, Hamid Beigy, Siavash Haghiri

On Mean Shift Clustering for Directional Data on a Hypersphere

The mean shift clustering algorithm is a useful tool for clustering numeric data. Recently, Chang-Chien et al. [1] proposed a mean shift clustering algorithm for circular data that are directional data on a plane. In this paper, we extend the mean shift clustering for directional data on a hypersphere.The three types of mean shift procedures are considered. With the proposed mean shift clustering for the data on a hypersphere it is not necessary to give the number of clusters since it can automatically find a final cluster number with good clustering centers. Several numerical examples are used to demonstrate its effectiveness and superiority of the proposed method.

Miin-Shen Yang, Shou-Jen Chang-Chien, Hsun-Chih Kuo

Backmatter

Weitere Informationen

Premium Partner

    Bildnachweise