Augmented model of stacked autoencoder for image classification

Abstract

Stacked Auto Encoder (SAE) is used to pre-train the deep network in the training phase of the individual layer for classifying complex real time data’s. MNIST and IMAGENET are used to train the network. Time consumed and accuracy during the training period is calculated for the MNIST data set which is binary image and IMAGENET dataset includes color image applying the Stacked Auto Encoder algorithm which is trained one layer at a time. Here the SAE consists of three layers which is stacked together and its parameters are varied in such a way that the constructed SAE out performs achieving time and accuracy tradeoff. The SAE model improves the accuracy of the image classifier in both binary and color image dataset with the reduced time.