Diverse Feature Learning by Self-distillation and Reset

Sejik Park

Our paper addresses the problem of models struggling to learn diverse features, due to either forgetting previously learned features or failing to learn new ones. To overcome this…