pings@fs0.ee.ubc.ca (pingnan shi) (10/23/90)
Has anyone made a comprehensive comparision between neural networks and digital parallel computers? What are the advantages of using neural networks on problems which can also be solved by parallel computers? There is no doubt that neural nets have their advantages when dealing with problems which have no analytic solutions, but how about problems which do have analytic solutions and we are only interested to solve them as fast as possible? Are neural networks faster than digital parallel computers, and if so how faster they can be theoratically? Do neural networks have the place of their own as a unique kind of parallel computers? These are some of the questions raised in my research, comments, references, and pointers are appreciated. Pingnan Shi Department of Electrical Engineering University of British Columbia Vancouver, B.C., Cananda E-mail: pings@ee.ubc.ca