Computed Decision Weights and a New Learning Algorithm for Neural Classifiers

TL;DR

This paper explores a novel approach to neural classifier training by computing decision weights directly through a new constrained optimization method, offering a simpler and effective alternative to traditional training.

Contribution

It introduces a new learning algorithm for neural classifiers that computes decision weights via constrained optimization, bypassing traditional training methods.

Findings

Proposes a new method for computing decision weights directly.

Demonstrates the effectiveness of the new learning process.

Offers a simpler alternative to conventional training techniques.

Abstract

In this paper we consider the possibility of computing rather than training the decision layer weights of a neural classifier. Such a possibility arises in two way, from making an appropriate choice of loss function and by solving a problem of constrained optimization. The latter formulation leads to a promising new learning process for pre-decision weights with both simplicity and efficacy.