Efficient Learning of Model Weights via Changing Features During Training

Marcell Beregi-Kovács, Ágnes Baran, András Hajdu

In this paper, we propose a machine learning model, which dynamically changes the features during training. Our main motivation is to update the model in a small content during the…