Utilization of Neural Network to Distinguish Mathematical Equations Patterns

Abstract

This paper deals with one of the most essential classification applications., it's the Pattern Recognition (PR) problem. The problem we want to manipulate in this paper is the recognition of simple mathematical equation pattern.

In this paper we present experiments comparing two different training algorithms, a Back Propagation (BP) and Particle Swarm Optimization (PSO) (as new algorithm to learn the NN), to train the artificial neural network (ANN), the training mean re-adjusting the weights of NN to increase the accuracy of results. This is accomplished by determining the fitness value, which is a threshold value.

This paper shows (probabilistically) which one of BP or PSO performs better than the other in simple mathematical equations recognition problem.