This book presents a systematic approach to parallel implementation of feedforward neural networks on an array of transputers. The emphasis is on backpropagation learning and training-set parallelism. Both theoretical analysis and experimental results for several bench mark problems are described. Guidelines for efficient parallel implementation are also highlighted. This book is suitable for graduate students, research scientists and practising engineers in the fields of artificial neural networks and parallel computing.

This volume consists of very high quality articles which may not only give a very good account of this field in the Soviet Union, but also provide stimulating materials for researchers working on this topic.

Concentrating mainly on the theorem of existence of periodic orbits for higher dimensional analogs of twist maps, this book is oriented somewhat toward the open problem of finding orbits of all (in particular, irrational) rotation vectors.

This text describes the basic methods of investigating convection patterns, and the types of two and three dimensional flows, pattern defects and sequences of convection-regime changes.