Catálogo de publicaciones - libros

The paper investigates the influence of versatile crossover and mutation operators on the efficiency of evolutionary search in solving two important classes of hard optimization problems. Chromosome representations of set partitions and permutations defined in the paper are not problem-oriented and are described together with their versatile variation operators. The proposed representations are tested in evolutionary programs on standard partitions and permutation problems i.e. graph coloring (GCP) and traveling salesman (TSP). The optimization results vary depending on the problem class. They are relatively positive with respect to GCP and negative for TSP.

A new evolutionary algorithm for induction of oblique decision trees is proposed. In contrast to the classical top-down approach, it searches for the whole tree at the moment. Specialized genetic operators are developed, which enable modifying both the tree structure and the splitting hyper-planes in non-terminal nodes. The problem of over-fitting can be avoided thanks to suitably defined fitness function. Experimental results on both synthetical and real-life data are presented and compared with obtained by the state-of-the-art decision tree systems.

Three nature-inspired algorithms are applied to solve Travelling Salesman Problem (TSP). The first originally developed Multi-agent Evolutionary Algorithm (MAEA) is based on multi-agent interpretation of TSP problem. An agent is assigned to a single city and builds locally its neighbourhood — a subset of cities, which are considered as local candidates to a global solution of TSP. Creating cycles — global solutions of TSP is based on Ant Colonies (AC) paradigm. Found cycles are placed in Global Table and are evaluated by genetic algorithm (GA) to modify a rank of cities in local neighbourhood. MAEA is compared with two another algorithms: artificial immune — based system (AIS) and a standard AC — both applied to TSP. We present experimental results showing that MAEA outperforms both AIS and AC algorithms.

Evolutionary algorithms work in an algorithmically simple manner but produce a huge amount of data. The extraction of useful information to gain further insight into the state of algorithm is a not-trivial task. In the paper, we propose a method of analysis of evolutionary algorithm by means of a graph theory. The method is inspired by latest results on scale-free network and small world phenomena. The paper presents visualization of evolutionary process based on network visualization software. The properties of such network are analyzed and various research possibilities are discussed.

The paper deals with the probability of misclassification in a multistage classifier. This classification problem is based on a decision-tree scheme. For given tree skeleton and features to be used, the Bayes decision rules at each non-terminal node are presented. Additionally the information on objects features is fuzzy or nonfuzzy. The upper bound of the difference between probability of misclassification for the both information’s is presented. In the paper we use the maximum likelihood estimator for fuzzy data.

In the paper, three conceptually simple but computer-intensive versions of an approach to selecting informative genes for classification are proposed. All of them rely on multiple construction of a tree classifier for many training sets randomly chosen from the original sample set, where samples in each training set consist of only a fraction of all of the genes. It is argued that the resulting ranking of genes can then be used to advantage for classification via a classifier of any type.

Balancing of parallel systems workload is very essential to ensure minimal response time of tasks submitted to process. Complexity of data warehouse systems is very high with respect to system structure, data model and many mechanisms used, which have a strong influence on overall performance. In this paper we present a dynamic algorithm of spatial telemetric data warehouse workload balancing. We implement HCAM data partitioning scheme which use Hilbert curves to space ordering. The scheme was modified in a way that makes possible setting of dataset size stored in each system node. Presented algorithm iteratively calculates optimal size of partitions, which are loaded into each node, by executing series of aggregation on a test data set. We investigate both situation in which data are and are not fragmented. Moreover we test a set of fragment sizes. Performed system tests conformed possibility of spatial telemetric data warehouse balancing algorithm realization by selection of dataset size stored in each node. Project was implemented in Java programming language with using of a set of available technology.

The problem of efficiency of general reasoning with knowledge updating on Bayesian networks is considered. The optimization should take into account not only the reasoning efficiency but also the prospective updating issues. An improved updating process based on an idea of data removing is proposed. Further possible improvement of the reasoning architecture is presented. Comparison of existing and proposed approaches are made on a basis of a computational experiment. Results of this experiment are presented.

The paper deals with the search and analysis of the subsequences in large volume sequences (texts, DNA sequences, etc.). A new algorithm ProMFS for mining frequent sequences is proposed and investigated. It is based on the estimated probabilistic-statistical characteristics of the appearance of elements of the sequence and their order. The algorithm builds a new much shorter sequence and makes decisions on the main sequence in accordance with the results of analysis of the shorter one.

In this paper we show that applying of multidimensional decompositions can improve the modelling results. The predictions usually consist of twofold elements, wanted and destructive ones. Rejecting of the destructive components should improve the model. The statistical methods like PCA and ICA with new modifications are employed. The example from the telecom market proofs correctness of the approach.