3 citations · 3 across the 2 of their papers we have counts for
1 paper · 1 filter
Brendan Ruff
Training a deep convolutional neural net typically starts with a random initialisation of all filters in all layers which severely reduces the forward signal and back-propagated er…