Study in image classification using nearest neighbors and perceptrons

Abstract

In this paper, we describe our experiments on training a computer to recognize (classify) a black and white image of hand-written digits. The experiments were done using Python. We try two methods to achieve the goal. The first method is k Nearest Neighbors, and the second is a Perceptron (modified to work for more than two classes). The paper describes each of these in detail, including some important implementation details, and reports results for these two methods, as well as some observations about their behavior.