Description
Principal components analysis (PCA) is a dimensionality reduction technique that allows to
compress high-dimensional data sets into very low dimensions.
Requirements:
Plot the 2000 MNIST digit images in Lab 8 to the 2 and 3 dimensional spaces respectively
after applying PCA. Also show how much variances of the data have been explained by the
principal components.
Sketch of the PCA algorithm:
Center your data
Compute the covariance matrix of centered matrix
Eigenvalue decomposition of covariance matrix
Project data into the low-dimensional space
