Abstract:
This note briefly discusses some of the classical results of McCulloch and Pitts. It then deals with some current research in neural nets. Several questions about neural nets are shown to be computationally difficult by showing that they are NP-Complete or worse. The size of neural nets necessary to compute functions or simulate machines is discussed, and some worst case optimal results are given. The statistics of randomly chosen autonomous nets are discussed, together with some approaches to understanding these statistics. Recent results on the behavior of analog nets are discussed, and the possibility of silicon implementation is mentioned.
