Neuro-dynamic programming. (English) Zbl 0924.68163

Belmont, MA: Athena Scientific. xiii, 491 p. (1996).
Topic of this book is what has been called formerly reinforcement learning, but what is now being addressed as neuro-dynamic programming. Three areas may be distinguished: Presentation of the background in dynamic programming, neural net architecture and stochastic approximation, secondly, the neuro-dynamic methodology proper newly developed by the authors, and finally the special case studies. One goal of the book is to proceed to rigorous proofs in an area where verbous arguments are omnipresent. This combination makes the book interesting: A mathematicians view on topics usually covered by imprecise qualitative reasoning. Some further catch words: finite/infinite horizon problems, stochastic shortest path problems, gradient methods in training, simulation of lookup table representation.


68T05 Learning and adaptive systems in artificial intelligence
68-01 Introductory exposition (textbooks, tutorial papers, etc.) pertaining to computer science