Parts-based learning by Non-Negative Matrix Factorisation

Visualising the principal components of portrait facial images. ‘Eigenfaces’ are the decomposed images in the direction of largest variance.Why we can’t relate to eigenfacesTraditional methods like Principal Component Analysis (PCA) would decompose a dataset into some form of latent representation e.g. eigenvectors, which at times can be meaningless when visualised — what actually is my first principal... Continue Reading →

Website Powered by

Up ↑