By Paul Smolensky,Michael C. Mozer,David E. Rumelhart

ISBN-10: 0805812016

ISBN-13: 9780805812015

ISBN-10: 1138876291

ISBN-13: 9781138876293

Mathematical types of neural networks reveal an grand richness and variety. Neural networks should be officially modeled as computational platforms, as actual or dynamical structures, and as statistical analyzers. inside of every one of those 3 large views, there are various specific ways. for every of sixteen specific mathematical views on neural networks, the contributing authors offer introductions to the historical past arithmetic, and handle questions such as:

* precisely what mathematical platforms are used to version neural networks from the given perspective?

* What formal questions about neural networks can then be addressed?

* What are common effects that may be received? and

* What are the phenomenal open problems?

a particular function of this quantity is that for every viewpoint provided in a single of the contributed chapters, the 1st editor has supplied a reasonably special precis of the formal effects and the considered necessary mathematical recommendations. those summaries are offered in 4 chapters that tie jointly the sixteen contributed chapters: 3 boost a coherent view of the 3 common views -- computational, dynamical, and statistical; the opposite assembles those 3 views right into a unified evaluate of the neural networks field.