A GA approach to Optimization of Convolution Neural Network

Abstract

In recent days a lot of activities in Deep Learning demonstrated ability to produce much better than other Machine Learning techniques. Much of the challenge in the Deep Learning is about optimizing the weights and several hyper parameters as it takes lot of computation and time to do. Gradient descent has been most popular technique currently in its weights optimization for back propagation. Most of the existing implementation of Convolution Neural Networks/Deep Learning Networks plays pivotal role in image processing. Though being scientifically regressive, BP and GD is slowly converging and getting easily trapped in local minima these are inherent disadvantages. For this reason, we explored another optimization with Meta Heuristic Algorithms such as Genetic Algorithm in the Deep Learning algorithm.