CODA: Constructivism Learning for Instance-Dependent Dropout Architecture Construction

Xiaoli Li

Dropout is attracting intensive research interest in deep learning as an efficient approach to prevent overfitting. Recently incorporating structural information when deciding whic…