Evolutionary approach for approximation of artificial neural network

Neural Network is an effective tool in the field of pattern recognition. The neural network classifies the pattern from the training data and recognizes if the testing data holds that pattern. The classical Back propagation (BP) algorithm is generally used to train the neural network for its simplicity. The basic drawback of this algorithm is its uncertainty and long training time and it searches the local optima and not the global optima. To overcome the drawback of Back propagation (BP) algorithm, here we use a hybrid evolutionary approach (GA-NN algorithm) to train neural networks. The aim of this algorithm is to find the optimized synaptic weight of neural network so as to escape from local minima and overcome the drawbacks of BP. The implementation is done taking images as input in ¿.png¿and ¿.tif¿ format.

I. INTRODUCTION

Evolutionary Algorithm (EA) and Neural Network (NN) represent two evolving technologies that are inspired by biological information science. NN is derived from brain theory to simulate learning behavior of an individual, that is, it is used for approximation (learning) and generalization (testing), while GA is developed from the evolutionary theory raised by Darwin to evolve the whole population for better fitness. Evolutionary Algorithm is actually used as an optimization algorithm, and not a learning algorithm. Optimization minimizes an error function. Minimization of the error will maximize generalization [3]. GA-NN algorithm has been used both to build and to train neural networks, electrical load forecasting-factor determination in columns etc.