Classification with SVD — Eigenfaces

Question

I just saw a play list on YouTube where Professor Steve Brunton teaches how the singular value decomposition (SVD) works and its applications, e.g., classification via the SVD. What is the difference between classification via SVD, say, [U,S,V] = svd(A), compared to the neural network $a = \sigma(Wx + b)$?

All I know that SVD have a "data bank", and neural networks...have not a "data bank" to compare with.

Question:

I know how to use classification for neural networks. But how can I do that for SVD?

Answer 1

SVD itself is a form of PCA.With EigenFaces you can decompose the Image into its constituents and use these vectors as features for an Classifier. So think of it as a feature extraction technique for images of faces .