A regularization technique that extends dropout to the weights of the network, setting a random subset of weights to zero during each training update. 27.07.2023 17:54 aior