An implementation of the k-NN learning rule as an algorithm for classification in Python. Includes both a naive and more optimized KD-tree variant.
Classification is performed on the MNIST dataset, consisting of handwritten digits in the form of images of 28 by 28 pixels withpixel intensities taking values in {0, 1, ... 255}. The images are accompanied by a ”true” label in {0, 1, ..., 9}. The data set is partitioned in training andtesting subsets and for the first couple of questions we will use a smaller sampleof the training and testing subsets to limit computation time.
For more details please see the whitepaper at https://github.com/792x/kNN/blob/master/kNN.pdf