They all rely on matrix multiplication ... loop which results in a constant improvement of the neural network in playing the game. Fawzi explained the novelty of AlphaTensor, and how it is able to ...