A demonstration of Principal Component Analysis (PCA) via the MNIST digit dataset and Pokemon images. Professor Peter Sadowski's PCA GitHub repo was used as a guideline for understanding PCA.
This notebook first gets the principal components of images of pokemon and reconstructs the images using their PCA eigenvectors. It then performs this same reconstruction on the MNIST digits dataset. Finally, transfer learning is performed. The pokemon images are embedded into the MNIST PCA space and mapped back into image space. This demonstrates the k value required to start to successfully reconstruct the pokemon images after the transfer learning.