Unsupervised Classifier for MNIST datasets

Overview

This is a basic Autoencoder that uses a two CNN's to create a latent feature space for the MNIST fashion dataset, though any MNIST data set can be used. It learns features about the data without ground truth classifiers, which can then be visualized via tSNE.