Miroslav Kubat

Gert Pfurtscheller

Irena Ivanova

Abstract: "This paper presents a new learning system and reports its first successful application to a medical domain. Polygraphic data were recorded during 8-hours sleep in 2 infants. A set of 15 parameters were estimated and submitted to an expert for classification. In this way, two data files were obtained and used as a test domain for the system. The tree-based neural net builds on the idea of generating a decision tree and translating it into a neural network architecture that is trainable by the backpropagation algorithm. The network has unambiguously defined topology and initial weights, and its subsequent training is very fast. The contribution of the system's individual aspects to the overall performance is studied. The achieved classification accuracy of cca 85% compares very favourably with other learning systems."

Tree-based Neural Net (TBNN) for Learning "difficult" Concepts

This textbook offers a comprehensive introduction to Machine Learning techniques and algorithms. This Third Edition covers newer approaches that have become highly topical, including deep learning, and auto-encoding, introductory information about temporal learning and hidden Markov models, and a much more detailed treatment of reinforcement learning. The book is written in an easy-to-understand manner with many examples and pictures, and with a lot of practical advice and discussions of simple applications. The main topics include Bayesian classifiers, nearest-neighbor classifiers, linear and polynomial classifiers, decision trees, rule-induction programs, artificial neural networks, support vector machines, boosting algorithms, unsupervised learning (including Kohonen networks and auto-encoding), deep learning, reinforcement learning, temporal learning (including long short-term memory), hidden Markov models, and the genetic algorithm. Special attention is devoted to performance evaluation, statistical assessment, and to many practical issues ranging from feature selection and feature construction to bias, context, multi-label domains, and the problem of imbalanced classes.

An Introduction to Machine Learning

Johannes Fürnkranz

The mind-set that has dominated the history of computer game playing relies on straightforward exploitation of the available computing power. The fact that a machine can explore millions of variations sooner than the sluggish human can wink an eye has inspired hopes that the mystery of intelligence can be cracked, or at least side-stepped, by sheer force. Decades of the steadily growing strength of computer programs have attested to the soundness of this approach. It is clear that deeper understanding can cut the amount of necessary calculations by orders of magnitude. The papers collected in this volume describe how to instill learning skills in game playing machines. The reader is asked to keep in mind that this is not just about games -- the possibility that the discussed techniques will be used in control systems and in decision support always looms in the background.

Miroslav Kubat

About

Books

Join the club