I have N rows of some training data, where each row is a $$(k_1,k_2,k_3)$$ dimensional matrix. The training data is in the form a Numpy Matrix, so a $$(N,k_1,k_2,k_3)$$ dime