Catálogo de publicaciones - libros

A novel neural filter and its application to active control of nonlinear noise are proposed in this paper. This method helps to avoid the premature saturation of backpropagation algorithm; meanwhile, guarantees the system convergence by the proposed self-tuning method. The comparison between the conventional filtered-X least-mean-square (FXLMS) algorithm and proposed method for nonlinear broadband noise in active noise cancellation (ANC) system is also made in this paper. The proposed design method is very easy to be implemented and versatile to the other applications. Several simulation results show that the proposed method can effectively cancel the narrowband and nonlinear broadband noise in a duct.

In this paper, it is shown how behavioural properties can be specified for three types of violent criminals. Moreover, it is shown how empirical material in the form of informal descriptions of traces of crime-related events can be formalised. Furthermore, it is shown how these formalised traces and behavioural properties can be used in automated analysis, for example in order to determine which type of criminal can have committed such a crime. Moreover, an underlying dynamical model is presented that shows causal mechanisms behind each of the behaviours, and their dependencies on the characteristics of the type of criminal and inputs in terms of stimuli from the environment.

Previous works have proposed to apply model-based diagnosis (MBD) techniques to detect and locate faults in the control software of mobile autonomous robots at runtime. The localization of faults at the level of software components enables the autonomous repair of the system by restarting failed components. Unfortunately, classical MBD approaches assume that components fail independently. In this paper we show that dependent failures are very common in this application domain and we propose the concept of in order to tackle the arising problems. We provide an algorithm for the computation of DEs and present the results of case studies.

It is generally accepted that knowledge based systems would be smarter if they can deal with uncertainty. Some research has been done to extend Description Logics(DLs) towards the management of uncertainty, most of which concerned the such as “The probability that a randomly chosen bird flies is greater than 0.9”. In this paper, we present a new kind of extended DLs to describe degrees of belief such as “The probability that all plastic objects float is 0.3”. We also introduce the extended tableau algorithm for Pr as an example to compute the probability of the implicit knowledge.

This study presents and evaluates a modified ant colony optimization (ACO) approach for the precedence and resource-constrained multiprocessor scheduling problems. A modified ant colony system, with two designed rules, called dynamic and delay ant colony system, is proposed to solve the scheduling problems. The dynamic rule is designed to modify the latest starting time of jobs and hence the heuristic function. A delay solution generation rule in exploration of the search solution space is used to escape the local optimal solution. Simulation results demonstrate that the proposed modified ant colony system algorithm provides an effective and efficient approach for solving multiprocessor system scheduling problems with precedence and resource constraints.

This study concerns the constructing of domain ontology from web tables in a specific domain. Ontology defines the common terms and their meaning (concepts) within a context. Thus only meaningful tables are our concern. The meaningful table is composed of a head and a body, which are formatted in rows and columns. The head abstracts the meaning expressed in the body. Thus, in order to obtain a table-information-extraction framework, this study extracts, as prerequisite work, the structural semantic, that is, the domain ontology that frames web-table information, from the head. We suggest a method for automatically extracting domain ontology using the structural and semantic characteristics of the web-table head. The construction of domain ontology proceeds through two steps: (a) extracting table schema as pseudo-ontology from each table from the same domain and (b) constructing domain ontology combining those extracted table schemata. The combination of schemata proceeds through splitting and clustering using (a) statistical information and (b) heuristics based on the structural and semantic characteristics of the web-table head.

The frequent pattern tree (FP-tree) is an efficient data structure for association-rule mining without generation of candidate itemsets. It, however, needed to process all transactions in a batch way. In the past, we proposed the Fast Updated FP-tree (FUFP-tree) structure to efficiently handle the newly inserted transactions in incremental mining. In this paper, we attempt to modify the FUFP-tree maintenance based on the concept of pre-large itemsets for efficiently handling deletion of records. Pre-large itemsets are defined by a lower support threshold and an upper support threshold. The proposed approach can thus achieve a good execution time for tree maintenance especially when each time a small number of records are deleted. Experimental results also show that the proposed Pre-FUFP deletion algorithm has a good performance for incrementally handling deleted records.

Selecting faculty members to form a mission committee and simultaneously scheduling the corresponding committee meetings is a tough decision problem frequently encountered in every academic department. In this paper, we present a formal model of the problem. We also present an approach showing that, with simple database construction, the problem can be transformed into a constrained itemset mining problem, which is an important branch of frequent itemset mining. Thus, the problem can be exactly solved by techniques for constrained itemset mining. For high efficiency, we provide a method to convert some problem constraint into an anti-monotone constraint, which can be easily embedded into the framework of frequent itemset mining and is considered very effective for search space pruning. Experiments were performed and the results show that our approach offers very high performance.

The learning algorithms of multilayered feed-forward networks can be classified into two categories, gradient and non-gradient kinds. The gradient descent algorithms like backpropagation (BP) or its variations are widely used in many application areas because of convenience. However, the most serious problem associated with the BP is local minima problem. We propose an improved gradient descent algorithm intended to weaken the local minima problem without doing any harm to simplicity of the gradient descent method. This algorithm is called dual gradient learning algorithm in which the upper connections (hidden-to-output) and the lower connections (input-to-hidden) separately evaluated and trained. To do so, the target values of hidden layer units are introduced to be used as evaluation criteria of the lower connections. Simulations on some benchmark problems and a real classification task have been performed to demonstrate the validity of the proposed method.

An adaptive algorithm for drifting environments is proposed and tested in simulated environments. Two powerful problem solving technologies namely Neural Networks and Genetic Algorithms are combined to produce intelligent agents that can adapt to changing environments. Online learning enables the intelligent agents to capture the dynamics of changing environments efficiently. The algorithm’s efficiency is demonstrated using a mine sweeper application. The results demonstrate that online learning within the evolutionary process is the most significant factor for adaptation and is far superior to evolutionary algorithms alone. The evolution and learning work in a cooperating fashion to produce best results in short time. It is also demonstrated that online learning is self sufficient and can achieve results without any pre-training stage. When mine sweepers are able to learn online, their performance in the drifting environment is significantly improved. Offline learning is observed to increase the average fitness of the whole population.