Pattern Recognition, International Conference on
Download PDF

Abstract

Nonrandom initial values of the weight vector can contain useful information. A weighted combination of the initial and final values of the weight vector can help to utilize this information. In adaptive training, at a start, we need to scale the initial weights and to stop training earlier before a minimum of the cost function is obtained. In order to weight, scale or to stop training optimally one needs to know accuracy of determination of initial and final weights.
Like what you’re reading?
Already a member?
Get this article FREE with a new membership!