Abstract: Neural networks (NN) often need to be lightweight and fast for practical deployment. Various NN compression techniques, including tensor decomposition, aim to achieve this goal. Tensor ...