Performance Enhancement Of Backpropagationalgorithm Using Momentum And Learningratewith A Case Study On Fingerprint Recognition

Abstract

Abstract— Backpropagation algorithm is very often used for learning process in multilayer networks of Artificial Neural Network (ANN) [7]. One area which oftenly uses ANN is fingerprint recognition. By applying learning rate [1] to backpropagation algorithm, learning process will be more stable and faster in finding the optimal delta (stepsize) on producing error . Momentum in backpropagation algorithm [2] monotonously decreases the errors during the training procedure with weakly convergent.

Combination of momentum and learningrate in the backpropagation algorithm during training process may increase the recognition accuracy of fingerprints. Our experiments show that by applying momentum and learningrate gradually to the backpropagation algorithm, recognition accuracy increased to 80.9%, which is 20% better than the standard backpropagation.

Keywords—Neural Networks; Fingerprint patterns; Backpropagation; Momentum; Learningrate