Catálogo de publicaciones - libros

The Fourth Workshop on Learning with Logics and Logics for Learning (LLLL2006) was held at June 5 and 6, 2006 in Tokyo, as one of the co-located workshops with the 20th Annual Conference of the Japanese Society for Artificial Intelligence (JSAI2006), and also as a meeting of the series SIG-FPAI of JSAI.

The workshop is proposed to bring together researchers who are interested in both of the areas of machine learning and computational logic, and to have intensive discussions on various relations between the two with making their interchange more active. We called for papers which are concerned with the following topics (but not exclusive): learning and knowledge discovery using logics, algorithmic aspects of learning based on logics, logics for machine learning and knowledge discovery, machine learning as a foundation of mathematics/ mathematical procedures, amalgamation of logic-based learning and statistical/ information theoretical learning, and learning and knowledge discovery from relational data, structured/semi-structured data or real-valued data. Also we organized the program committee for the workshop consisting of 16 excellent researchers to the area in logic and/or learning.

A consistent learner is required to correctly and completely reflect in its actual hypothesis all data received so far. Though this demand sounds quite plausible, it may lead to the unsolvability of the learning problem.

Therefore, in the present paper several variations of consistent learning are introduced and studied. These variations allow a so-called –delay relaxing the consistency demand to all but the last data.

Additionally, we introduce the notion of learning (again with –delay) requiring the learner to correctly reflect only the last datum (only the  − th datum) seen.

Our results are threefold. First, it is shown that all models of coherent learning with –delay are exactly as powerful as their corresponding consistent learning models with –delay. Second, we provide characterizations for consistent learning with –delay in terms of complexity. Finally, we establish strict hierarchies for all consistent learning models with –delay in dependence on .

In this paper, we generalize previous results showing connections between inductive inference from positive data and algebraic structures by using tools from universal algebra. In particular, we investigate the inferability from positive data of language classes defined by closure operators. We show that some important properties of language classes used in inductive inference correspond closely to commonly used properties of closed set systems. We also investigate the inferability of algebraic closed set systems, and show that these types of systems are inferable from positive data if and only if they contain no infinite ascending chain of closed sets. This generalizes previous results concerning the inferability of various algebraic classes such as the class of ideals of a ring. We also show the relationship with algebraic closed set systems and approximate identifiability as introduced by Kobayashi and Yokomori [11]. We propose that closure operators offer a unifying framework for various approaches to inductive inference from positive data.

This paper presents a branch and bound search algorithm for finding only top number of extents of formal concepts w.r.t. their evaluation, where the corresponding intents are under some quality control. The algorithm aims at finding potentially interesting documents of even lower evaluation values that belong to some highly evaluated formal concept. The experimental results show that it can effectively find such documents.

In the present paper, we propose a new method of -gram analysis using ZBDDs (Zero-suppressed BDDs). ZBDDs are known as a compact representation of combinatorial item sets. Here, we newly apply the ZBDD-based techniques for efficiently handling sets of sequences. Using the algebraic operations defined over ZBDDs, such as union, intersection, difference, etc., we can execute various processings and/or analyses for large-scale sequence data. We conducted experiments for generating -gram statistical data for given real document files. The obtained results show the potentiality of the ZBDD-based method for the sequence database analysis.

International workshop on Risk Mining (RM2006) was held in conjunction with the 20th Annual Conference of the Japanese Society for Artificial Intelligence(JSAI2005), Tokyo Japan, June 2005. The workshop aimed at sharing and comparing experiences on risk mining techniques applied to risk detection, risk clarification and risk utilization. In summary, the workshop gave a discussion forum for researchers working on both data mining and risk management where the attendees discussed various aspects on data mining based risk management.

The knowledge on the relation between a financial state of an enterprise and its future profit will efficiently and securely reduce the negative risk and increase the positive risk on the decision making needed in the management of the enterprise and the investment in stock markets. Generally speaking, the relation is considered to have a highly complicated structure containing the influences from various financial factors characterizing the enterprise. Recent development of data mining techniques has significantly extended the power to model such a complicated relation in accurate and tractable manners. In this study, we assessed the feasibility to model the relation in the framework of data mining, and analyzed the characteristics of the model.

Kyushu Electric Power Co.,Inc. collects different sensor data and weather information to maintain the safety of hydroelectric power plants while the plants are running. In order to maintain the safety of hydroelectric power plants, it is very important to measure and collect the sensor data of abnormal condition and trouble condition. However, it is very hard to measure and collect them. Because it is very rare to occur abnormal condition and trouble condition in the hydroelectric power equipment. In this situation, we have to find abnormal condition sign as a risk management from the many sensor data of normal condition. In this paper, we consider that the abnormal condition sign may be unusual condition. This paper shows results of unusual condition patterns detection of bearing vibration. The unusual condition patterns are detected from the collected different sensor data and weather information by using one class support vector machine. The result shows that our approach may be useful for unusual condition patterns detection in bearing vibration and maintaining hydroelectric power plants. Therefore, the proposed method is one method of risk management for hydroelectric power plants.

This paper presents a nonlinear structural health assessing technique, based on an interactive data mining approach. A data mining control agent emulating cognitive process of human analyst is integrated in the data mining loop, analyzing and verifying the output of the data miner and controlling the data mining process to improve the interaction between human user and computer system. Additionally, an artificial neural network method, which is adopted as a core component of the proposed interactive data mining method, is evolved by adding a novelty detecting and retraining function for handling complicated nuclear power plant quake-proof data. To demonstrate how the proposed technique can be used as a powerful tool for assessment of structural status in nuclear power plant, quake-proof testing data has been applied.

E-finance industry is rapidly transforming and evolving toward more dynamic, flexible and intelligent solutions. This paper describes a model with dynamic multi-level workflows corresponding to a multi-layer Grid architecture, for multi-aspect analysis in building e-finance portals on the Wisdom Web. The application and research demonstrate that mining-grid centric three-layer Grid architecture is effective for developing intelligent risk management and decision making financial systems.

This paper concentrates on how to develop an mining-grid centric e-finance portal (MGCFP), not only for supplying effective online financial services for both retail and corporate customers, but also for intelligent credit risk management and decision making for financial enterprises and partners.