Table of Contents

Preface. Acknowledgments. List of Figures. List of Tables. List of Contributors. INTRODUCTION. Dynamical Recurrent Networks (J. Kolen and S. Kremer). ARCHITECTURES. Networks with Adaptive State Transitions (D. Calvert and S. Kremer). Delay Networks: Buffers to Rescue (T. Lin and C. Giles). Memory Kernels (A. Tsoi, et al.). CAPABILITIES. Dynamical Systems and Iterated Function Systems (J. Kolen). Representation of Discrete States (C. Giles and C. Omlin). Simple Stable Encodings of Finite-State Machines in Dynamic Recurrent Networks (M. Forcada and R. Carrasco). Representation Beyond Finite States: Alternatives to Pushdown Automata (J. Wiles, et al.). Universal Computation and Super-Turing Capabilities (H. Siegelmann). ALGORITHMS. Insertion of Prior Knowledge (P. Frasconi, et al.). Gradient Calculations for Dynamic Recurrent Neural Networks (B. Pearlmutter). Understanding and Explaining DRN Behavior (C. Omlin). LIMITATIONS. Evaluating Benchmark Problems by Random Guessing (J. Schmidhuber, et al.). Gradient Flow in Recurrent Nets: The Difficulty of Learning Long-Term Dependencies (S. Hochreiter, et al. Limiting the Computational Power of Recurrent Neural Networks: VC Dimension and Noise (C. Moore). APPLICATIONS. Dynamical Reccurent Networks in Control (D. Prokhorov, et al.). Sentence Processing and Linguistic Structure (W. Tabor). Neural Network Architectures for the Modeling of Dynamic Systems (H. Zimmerman and R. Neuneier). From Sequences to Data Structures: Theory and Applications (P. Frasconi, et al.). CONCLUSION. Dynamical Recurrent Networks: Looking Back and Looking Forward (S. Kremer and J. Kolen). Bibliography. Glossary. Index. About the Editors.

About the Author

About the Editors John F. Kolen has explored the computational capabilities of dynamical recurrent networks on a wide range of projects: computer tomography of ballistic tests, autonomous science on extraterrestrial sensor platforms, and laser marksmanship modeling. His research interests include neural networks, distributed processing, philosophy of computation, and computer gaming. Dr. Kolen is a member of the Institute for Human and Machine Cognition at the University of West Florida. Stefan C. Kremer's research interests include connectionist networks (the subject of his 1996 thesis "A Theory of Grammatical Induction in the Connectionist Paradigm"), genetic algorithms, signal processing, grammar induction, and image processing. He is an assistant professor of computing and information science at the University of Guelph, Ontario, Canada, and is a founding member of the Guelph Natural Computation Research Group.