Design and Implementation of ConvNet for Handwritten Digits Classification on Graphical Processing Unit

Convolutional Neural Network (CNN) or ConvNet is the leading-edge deep learning model that has achieved phenomenal successes in the tasks of image classification, object recognition, speech recognition and natural language processing. ConvNets are inherently complex architectures and training ConvNets requires significant amount of computation. There is a need to determine whether GPU or CPU provide the effective method for implementation of ConvNets. Very few studies have come up with the comparison of the implementation of ConvNets on both CPU and GPU. The present work examines the impact of GPU on the implementation of ConvNets. ConvNet is trained on MNIST dataset to perform classification of handwritten digits. The experiments have been performed on both CPU and GPU and observed that there is a performance improvement of 5 times in terms of training time speedup on GPU. The proposed work also investigates the effect of regularization and the results show that regularization indeed reduces the problem of overfitting.