Learning through atypical ''phase transitions'' in overparameterized neural networks

Open in new window