I would like to train a network in such a way that it is composed of factors of the overall network. When i pass a value through it , i get a 512 dimensional vector as outpu
