Adaptive Processing of Sequences and Data Structures

This book is devoted to adaptive processing of structured information similar to flexible and intelligent information processing by humans - in contrast to merely sequential processing of predominantly symbolic information within a deterministic framework. Adaptive information processing allows for a mixture of sequential and parallel processing of symbolic as well as subsymbolic information within deterministic and probabilistic frameworks.
The book originates from a summer school held in September 1997 and thus is ideally suited for advanced courses on adaptive information processing and advanced learning techniques or for self-instruction. Research and design professionals active in the area of neural information processing will find it a valuable state-of-the-art survey.

The book originates from a summer school held in September 1997 and is thus ideally suited for advanced courses on adaptive information processing and advanced learning techniques or for self-instruction. Includes supplementary material: sn.pub/extras

Inhalt
Recurrent neural network architectures: An overview.- Gradient based learning methods.- Diagrammatic methods for deriving and relating temporal neural network algorithms.- An introduction to learning structured information.- Neural networks for processing data structures.- The loading problem: Topics in complexity.- Learning dynamic Bayesian networks.- Probabilistic models of neuronal spike trains.- Temporal models in blind source separation.- Recursive neural networks and automata.- The neural network pushdown automaton: Architecture, dynamics and training.- Neural dynamics with stochasticity.- Parsing the stream of time: The value of event-based segmentation in a complex real-world control problem.- Hybrid HMM/ANN systems for speech recognition: Overview and new research directions.- Predictive models for sequence modelling, application to speech and character recognition.