mnist

Open University Computer Vision 22928- Ex1 Classical Machine Learning Algorithms with MNIST.

Full report can be found under /report/report.docx

KNN

Calculate the performance of the KNN classifier, for k=1..10.

PCA

Calculate PCA on MNIST.
Draw the average digit, and the 6 first principal components.
Draw a graph of the explained variance, by n most significant components.
How many components are required to get to 95% variance? 80%?
Project the digits to dimension 2, and draw the obtained vectors, each digit with its own color.
Repeat the KNN question, where each digit is represented by its projection to dimension 2, 10, 20.
For some digit, project it to dimension k, then restore it. Draw the restoration for different k.
Calculate PCA for every digit separately. Present the 1st 6 principal components of each.
Calculate the projection of each test set image to each model.
Restore by each model.
Calculate the distance from each restoration to the original image.
Select the model for which the distance is smallest.