Asymptotics of feature learning in two-layer networks after one gradient-step

Open in new window