The trick is to transform the convolution step in a convolutional neural network into a series of matrix multiplications.

Repeated convolution and truncation of a truncated fat-tailed distribution, instead of Monte Carlo simulation, for pricing a ...