1 paper
Rafał Surdej, Michał Bortkiewicz, Alex Lewandowski +2
Trainable activation functions, whose parameters are optimized alongside network weights, offer increased expressivity compared to fixed activation functions. Specifically, trainab…