Resnet in Resnet: Generalizing Residual Architectures

Sasha Targ, Diogo Almeida, Kevin Lyman

Abstract: ResNets have recently achieved state-of-the-art results on challenging computer vision tasks. In this paper, we create a novel architecture that improves ResNets by adding the ability to forget and by making the residuals more expressive, yielding excellent results. ResNet in ResNet outperforms architectures with similar amounts of augmentation on CIFAR-10 and establishes a new state-of-the-art on CIFAR-100.

Conflicts: enlitic.com, ucsf.edu