1 paper
Pawel Pukowski, Haiping Lu
Despite their limited interpretability, weights and biases are still the most popular encoding of the functions learned by ReLU Neural Networks (ReLU NNs). That is why we introduce…