Investigate GPU benefits for linear algebra #41
Description
As we are limited by tasks that can be parallelized embarassingly (e.g. Phi[i,j] := Phi[n,p,i] * Phi[n,p,j]) and also ensemble method loops in e.g. the hyperparameter training. It may be useful to investigate GPU accelerations