Introduction to Neural Network and Improved Algorithm to Avoid Local Minima and Faster Convergence

Abstract

Neural network is a massively parallel distributed processor inspired by the structure and functional aspect of biological neural network that employ learning algorithm, like back propagation for computation .back propagation follows supervised learning rule for processing data. The architecture is complex and processing confronts problems like local minima, slow convergence and premature saturation. We have introduced improved neuron model and learning rule like multiplicative neuron model using extra term in algorithm called a proportional factor. This along with other modifications in network architecture, that helps attain faster learning and accurate computation.KeywordsSupervised LearningBack Propagation AlgorithmPremature SaturationDifferential Adaptive LearningMultiplicative Neuron Model