Catálogo de publicaciones - libros

In this paper we present a novel data mining technique, known as Post Sequential Patterns Mining, which can be used to discover Structural Patterns. A Structural Pattern is a new pattern, which is composed of sequential patterns, branch patterns or iterative patterns. Sequential patterns mining plays an essential role in many areas and substantial research has been conducted on their analysis and applications. In our previous work [12], we used a simple but efficient Sequential Patterns Graph (SPG) to model the sequential patterns. The task to discover hidden Structural Pattern is based on our previous work and sequential patterns mining, conveniently named Post Sequential Patterns Mining. In this paper, in addition to stating this new mining problem, we define patterns such as branch pattern, iterative pattern, structural pattern, and concentrate on finding concurrent branch pattern. Concurrent branch pattern is thus one of the main forms of structural pattern and will play an important role in event-based data modelling.

CP-networks are an elegant and compact qualitative framework for express preference, in which we can represent and reason about preference rankings given conditional preference statements. However, represent constraints in such framework is one difficult problem. We therefore propose a new approach, i.e. mapping CP-networks to constraint hierarchy, thus we can reason preferences with constraint solving algorithms. We compare it with related work finally.

We proposed a neural network called LPPH-CSP for solving constraint satisfaction problem (CSP). The LPPH-CSP is not trapped by any point which is not a solution of the CSP, and it can update all neurons simultaneously. In this paper, we propose two methods to improve the efficiency of the LPPH-CSP. Though the LPPH-CSP can deal with several types of constraints of the CSP, it treats all constraints evenly. One of the proposed methods distinguishes the types of constraints for solving the CSP more efficiently. Another one of the proposed methods applies fast local search (FLS) to the LPPH-CSP. Experimental results show the effectiveness of our proposals.

Component retrieval, about how to locate and identify appropriate components, is one of the major problems in component reuse. It becomes more critical as more reusable components come from component markets instead of from an in-house component library, and the number of available components is dramatically increasing. In this paper, we review the current component retrieval methods and propose our conversational component retrieval model (CCRM). In CCRM, components are represented as cases, a knowledge-intensive case-based reasoning (CBR) method is adopted to explore context-based semantic similarities between users’ query and stored components, and a conversational case-based reasoning (CCBR) technology is selected to acquire users’ requirements interactively and incrementally.

LPPH has been proposed to solve the satisfiability problem (SAT). In order to solve the SAT more efficiently, a parallel execution has been proposed. Experimental results show that higher speedup ratio is obtained by using this parallel execution of the LPPH. In this paper, we propose a method of mixed parallel execution of several algorithms for the SAT. “Mixed” means the parallel execution of the LPPH and local search algorithms. In the experiments, we used the LPPH with attenuation coefficient generating function and the GSAT. Results of experiments show mixing these two algorithms yield excellent performance.

We study four kinds of binary codes of amino acids (AA). Two codes of them are based respectively on biochemical properties, and the two others are generated with artificial intelligence (AI) methods, and are based on protein structures and alignment, and on Dayhoff matrix. In order to give a global significance of each binary code, we use a hierarchical clustering method to generate different clusters of each binary codes of amino acids. Each cluster is examined with biochemical properties to give an explanation on the similarity between amino acids that it contains. To validate our examination, a decision tree based machine learning system is used to characterize the AA clusters obtained with each binary codes. From this experimentation, it comes out that one of the AI based codes allows to obtain clusters that have significant biochemical properties. As a consequence, it appears that even if attributes of binary codes generated with AI methods, do not separately correspond to a biochemical property, they can be significant in the whole. Conversely binary codes based on biochemical properties can be insignificant when forming a whole.

Natural language interface (NLI) facilitates the human use of computers. In this paper, we review the state-of-the-art NLI application. Based on the extant literature, we design process flow of an NLI system enabling easy information access via mobile devices.

The paper analyzes some existing incremental mining algorithms for association rule and presents an incremental mining algorithm for association rule fit for Web Usage Mining. Because there are some characteristics of web logs which are dynamic, attributed, smaller and updated frequently, the algorithm uses BORDERS algorithm when mining single log file, and takes advantage of partition algorithm when mining many log files simultaneously.

In this paper, a face recognition algorithm based on modular ICA approach is presented. Compared whit conventional ICA algorithm, the proposed algorithm has an improved recognition rate for face images with large variations in lighting direction and facial expression. In the proposed technique, the face images are divided into smaller sub-images and the ICA approach is applied to each of these sub-images. Since some of the local facial features of an individual do not vary even when the pose, lighting direction and facial expression vary, we expect the proposed method to be able to cope with these variations. The accuracy of the conventional ICA method and modular ICA method are evaluated under the conditions of varying expression, illumination and pose using Yale face database[1].

Recent years have seen considerable developments in modeling techniques for automatic fault location in programs. However, much of this research considered the models from a standalone perspective. Instead, this paper focuses on the highly unusual properties of the testing and measurement process, where capabilities differ strongly from the classical hardware diagnosis paradigm. In particular, in an interactive debugging process user interaction may result in highly complex input to improve the process. This work extends the standard entropy-based measurement selection algorithm proposed in (de Kleer and Williams, 1987) to deal with high-level observations about the intended behavior of Java programs, specific to a set of test cases. We show how to incorporate the approach into previously developed model-based debugging frameworks and to how reasoning about high-level properties of programs can improve diagnostic results.