Machine Learning Approaches for Face Identification Feed Forward Algorithms

Abstract

Face identification using feed forward technique is a very important technique to use in computer vision, machine learning, biometrics, pattern recognition, pattern analysis and digital image processing. It is a systematic method for training multi-layer convolutional neural network. It is a mathematical technique that is strong but not highly used in practical. Feed forward technique is using for extend gradient descent based delta learning rules. Feed forward technique are provides a computationally efficient method for changing the weight and bias. Face learning problem is to search for all hypothesis space defined to all weight values for all units in the networks. The error is replaced by P and the other category of the space corresponding to all of the associated weight with all of the units in the network. In this equation in the case of training a single unit the output attempts to find a hypothesis to minimize P. In face identification algorithm the automatically determined location of the different feature. This alignment is refined by optical view. Identification is performing by computing normalized correlation scores in many face identification scenarios the pose of the probe and registered database image are different.