camargo@cs.columbia.edu (Kiko) (12/15/90)
Hi there, I just completed a final draft of a summary paper on learning algorithms for Neural Networks. I'd like to put it at your disposal, and I welcome your comments. Please, forgive me for eventual typos, and help me fixing them if you can. In case you find real errors, please, don't hesitate in pointing them out. This is my first attempt in this kind of endeavor, and I'd like to get it right. The paper covers Associative Nets (Anderson, Hopfield, etc) with a little about higher order nets. Then goes on to Boltzman Machines, Feed Forward Nets, Back Propagation, G-Maximization, Feature Maps, ART and a little on reinforcement learning. I don't cover the Conjugate Gradient methods, nor Genetic Algorithms, but I think the paper makes a good introduction for the uninitiated. The paper is 52 pages long, with a bibliography of 86 entries. If you are interested, you can get a copy through FTP by connecting into "amazonas.cs.columbia.edu", or internet address 128.59.16.72. Login as anonymous, and use your "id" as password. Change directory into "/pub". Set mode to "binary" and "get learning.ps". If you can't get it electronically, send mail to "camargo@cs.columbia.edu" and I'll see that you get a printed copy. But I won't be able to send it before the end of january, for I'm leaving the country in the next few days. I really appreciate any comments. Thanks, Francisco A. Camargo camargo@cs.columbia.edu