Contrastive Examination of Learning Techniques Deployed Over Handwritten Digits

Abstract

In today’s world, great leaps and advances have been achieved in the sphere of machine learning, specifically in the sub-sphere of character recognition. In this paper we perform a contrastive examination to compare three major classification algorithms on the basis of accuracy and training times. The techniques were namely, K-Nearest Neighbors (KNN), Convolutional Neural Networks (CNN) and Support Vector Machines (SVMs) with their respective accuracies being 97.85%, 99.93% and 99.96%. They were applied on well-know and well curated MNIST Dataset maintained by Courant Institute - NYU, Google Labs - New York and Microsoft Research - Redmond.