PCA LDA FDA

This repository contains python implementation of PCA, LDA and FDA from scratch on MNIST dataset. Noise removal effect of PCA is also depicted visually.

Dataset used:

MNIST

Steps taken:

1. Finding the global mean and covariance of data.
2. Implementation of PCA, FDA and LDA from scratch.
3. Visualize and analyze the eigenvectors obtained using PCA with 95% eigen energy.
4. Finding accuracies at different values of eigen energies.
5. Displaying eigenvectors by converting them into image form
6. Performing FDA then LDA.
7. Performing PCA then FDA.

Instructions

Just enter path to the idx train test files in the code.