1 paper
Skander Karkar, Ibrahim Ayed, Emmanuel de Bézenac +1
Greedy layer-wise or module-wise training of neural networks is compelling in constrained and on-device settings where memory is limited, as it circumvents a number of problems of…