Identifying digits in the MNIST dataset using Unsupervised Learning with K-means, Autoencoding, and PCA; and Denoising on the MNIST dataset
This project includes experiments showing that classification using K-means on Principal Component Analysis-reduced encodings performs better than K-means on encodings alone.
A second experiment shows the effectiveness of denoising using an Autoencoder.
The notebook for this project is here.
The report is included as a pdf in this repository.