Information Theoretic Neural Computation

A detailed formulation of neural networks from the information-theoretic viewpoint. The authors show how this perspective provides new insights into the design theory of neural networks. In particular they demonstrate how these methods may be applied to the topics of supervised and unsupervised learning, including feature extraction, linear and non-linear independent component analysis, and Boltzmann machines. Readers are assumed to have a basic understanding of neural networks, but all the relevant concepts from information theory are carefully introduced and explained. Consequently, readers from varied scientific disciplines, notably cognitive scientists, engineers, physicists, statisticians, and computer scientists, will find this an extremely valuable introduction to this topic.

In order to develop new types of information media and technology, it is essential to model complex and flexible information processing in living systems. This book presents a new approach to modeling complex information processing in living systems. Traditional information-theoretic methods in neural networks are unified in one framework, i.e. α-entropy. This new approach will enable information systems such as computers to imitate and simulate human complex behavior and to uncover the deepest secrets of the human mind. Contents: Information in Neural NetworksInformation MinimizationInformation MaximizationConstrained Information MaximizationNeural Feature DetectorsInformation Maximization and MinimizationInformation ControllerInformation Control by α-EntropyIntegrated Information Processing Systems Readership: Students and researchers in artificial intelligence and neural networks. Keywords:

Comprehensive introduction to the neural network models currently under intensive study for computational applications. It also provides coverage of neural network applications in a variety of problems of both theoretical and practical interest.

Information theoretics vis-a-vis neural networks generally embodies parametric entities and conceptual bases pertinent to memory considerations and information storage, information-theoretic based cost-functions, and neurocybernetics and self-organization. Existing studies only sparsely cover the entropy and/or cybernetic aspects of neural information. Information-Theoretic Aspects of Neural Networks cohesively explores this burgeoning discipline, covering topics such as: Shannon information and information dynamics neural complexity as an information processing system memory and information storage in the interconnected neural web extremum (maximum and minimum) information entropy neural network training non-conventional, statistical distance-measures for neural network optimizations symmetric and asymmetric characteristics of information-theoretic error-metrics algorithmic complexity based representation of neural information-theoretic parameters genetic algorithms versus neural information dynamics of neurocybernetics viewed in the information-theoretic plane nonlinear, information-theoretic transfer function of the neural cellular units statistical mechanics, neural networks, and information theory semiotic framework of neural information processing and neural information flow fuzzy information and neural networks neural dynamics conceived through fuzzy information parameters neural information flow dynamics informatics of neural stochastic resonance Information-Theoretic Aspects of Neural Networks acts as an exceptional resource for engineers, scientists, and computer scientists working in the field of artificial neural networks as well as biologists applying the concepts of communication theory and protocols to the functioning of the brain. The information in this book explores new avenues in the field and creates a common platform for analyzing the neural complex as well as artificial neural networks.

Table of contents

Information theoretics vis-a-vis neural networks generally embodies parametric entities and conceptual bases pertinent to memory considerations and information storage, information-theoretic based cost-functions, and neurocybernetics and self-organization. Existing studies only sparsely cover the entropy and/or cybernetic aspects of neural information. Information-Theoretic Aspects of Neural Networks cohesively explores this burgeoning discipline, covering topics such as: Shannon information and information dynamics neural complexity as an information processing system memory and information storage in the interconnected neural web extremum (maximum and minimum) information entropy neural network training non-conventional, statistical distance-measures for neural network optimizations symmetric and asymmetric characteristics of information-theoretic error-metrics algorithmic complexity based representation of neural information-theoretic parameters genetic algorithms versus neural information dynamics of neurocybernetics viewed in the information-theoretic plane nonlinear, information-theoretic transfer function of the neural cellular units statistical mechanics, neural networks, and information theory semiotic framework of neural information processing and neural information flow fuzzy information and neural networks neural dynamics conceived through fuzzy information parameters neural information flow dynamics informatics of neural stochastic resonance Information-Theoretic Aspects of Neural Networks acts as an exceptional resource for engineers, scientists, and computer scientists working in the field of artificial neural networks as well as biologists applying the concepts of communication theory and protocols to the functioning of the brain. The information in this book explores new avenues in the field and creates a common platform for analyzing the neural complex as well as artificial neural networks.

Written by the three leading authorities in the field, this book brings together -- in one volume -- the recent developments in discrete neural computation, with a focus on neural networks with discrete inputs and outputs. It integrates a variety of important ideas and analytical techniques, and establishes a theoretical foundation for discrete neural computation. Discusses the basic models for discrete neural computation and the fundamental concepts in computational complexity; establishes efficient designs of threshold circuits for computing various functions; develops techniques for analyzing the computational power of neural models. A reference/text for computer scientists and researchers involved with neural computation and related disciplines.

In this richly illustrated book, it is shown how Shannon's mathematical theory of information defines absolute limits on neural efficiency; limits which ultimately determine the neuroanatomical microstructure of the eye and brain. Written in an informal style this is an ideal introduction to cutting-edge research in neural information theory.