1 paper
Mohamad Mohamad, Francesco Ponzio, Xavier Descombes
Mode-dependent architectural components (layers that behave differently during training and evaluation, such as Batch Normalization or dropout) are commonly used in visual reinforc…