This paper puts forward a method using neural networks to minimize the Wasserstein distance between log signatures to reduce the target dimension of the ...