rubicco/mnist-semi-supervised

Semi-supervised learning on MNIST digits dataset with very limited amount of label (10; 1-per-class). A simple AutoEncoder model is built to learn input representations and KNN is used to achieve classification.

Jupyter NotebookMIT