Polynomial neural network features MNIST

Question

i'm trying to create a neural network with MNIST data set. I know I have to unroll each image into a vector with every pixel, and stack each image vector, but how do I add polynomial features? I am doing this in python

Answer 1

Vector is rarely use in digit classification, people use numpy array instead. As image pixels, Weights and biases are also stored in numpy arrays, and we do the further actions such as multiply, which is probably your "polynomial features" means, on numpy arrays as in matrices. There is plenty of tutorial that you can search for but I would recommend this.