Source: https://ai.stanford.edu/~ang/papers/nips09-AudioConvolutionalDBN.pdf