tag:blogger.com,1999:blog-7994087232040033267.post7500304908818237398..comments2024-03-03T00:23:26.457-08:00Comments on Pragmatic Programming Techniques: Machine Learning with Linear ModelRicky Hohttp://www.blogger.com/profile/03793674536997651667noreply@blogger.comBlogger1125tag:blogger.com,1999:blog-7994087232040033267.post-33784030905320453782009-12-02T00:39:01.999-08:002009-12-02T00:39:01.999-08:00Hi Mr. Ho, first of all congratulation for your bl...Hi Mr. Ho, first of all congratulation for your blog. It's so interesting! :)<br /><br />In respect of Neural Networks, there are different tecniques we can use to adjust weights and they substantially depend from neural network topology.<br /><br />In particular there are:<br /> - Wiener-Hopf Method<br /> - Steepest Descent Method (your Gradient Algorithm)<br /> - Least-Mean-Square Algorithm (Stocastic Gradient Algorithm)<br /><br />With Kolmogorov theorem we know that each continue, limited and mototone function (with n variables) could be represented as sum of many mono-variable functions. The problem is that this is a Theorem of existence: it say that exists a set of functions but do not say how we can calculate them. For this reason we use a Neural Network.<br /><br />Bye and congratulation again.Anonymousnoreply@blogger.com