New Learning Paradigms in Soft Computing

Learning is a key issue in the analysis and design of all kinds of intelligent systems. In recent time many new paradigms of automated (machine) learning have been proposed in the literature. Soft computing, that has proved to be an effective and efficient tool in so many areas of science and technology, seems to offer new qualities in the realm of machine learning too. The purpose of this volume is to present some new learning paradigms that have been triggered, or at least strongly influenced by soft computing tools and techniques, mainly related to neural networks, fuzzy logic, rough sets, and evolutionary computations.

Statistical Learning by Natural Gradient Descent.- Granular Networks and Granular Computing.- Learning and Decision-Making in the Framework of Fuzzy Lattices.- Lazy Learning: A Logical Method for Supervised Learning.- Active Learning in Neural Networks.- Knowledge Extraction from Reinforcement Learning.- Reinforcement Learning for Fuzzy Agents: Application to a Pighouse Environment Control.- Performance Comparisons of Neural Networks and Machine Learning Techniques: A Critical Assessment of the Methodology.- Digital Systems Design Through Learning.- Hybrid Inductive Machine Learning: An Overview of CLIP Algorithms.- An Integer Programming Approach to Inductive Learning Using Genetic and Greedy Algorithms.- Using Unlabeled Data for Learning Classification Problems.- Problems of Rule Induction from Preterm Birth Data.- Reduction of Discriminant Rules Based on Frequent Item Set Calculation.- Deriving a Concise Description of Non-Self Patterns in an Artificial Immune System.