Scaling the Scattering Transform: Deep Hybrid Networks
This work addresses the need for more efficient and data-efficient deep learning models, particularly in scenarios with limited data, though it is incremental as it builds on existing scattering and hybrid network methods.
The paper tackles the problem of improving deep network efficiency by using a scattering transform as a fixed initialization for early layers, achieving AlexNet accuracy on ImageNet with a shallow cascade and a top-5 error of 11.4% when combined with ResNet, while also showing better performance in small sample regimes on CIFAR-10 and STL-10.
We use the scattering network as a generic and fixed ini-tialization of the first layers of a supervised hybrid deep network. We show that early layers do not necessarily need to be learned, providing the best results to-date with pre-defined representations while being competitive with Deep CNNs. Using a shallow cascade of 1 x 1 convolutions, which encodes scattering coefficients that correspond to spatial windows of very small sizes, permits to obtain AlexNet accuracy on the imagenet ILSVRC2012. We demonstrate that this local encoding explicitly learns invariance w.r.t. rotations. Combining scattering networks with a modern ResNet, we achieve a single-crop top 5 error of 11.4% on imagenet ILSVRC2012, comparable to the Resnet-18 architecture, while utilizing only 10 layers. We also find that hybrid architectures can yield excellent performance in the small sample regime, exceeding their end-to-end counterparts, through their ability to incorporate geometrical priors. We demonstrate this on subsets of the CIFAR-10 dataset and on the STL-10 dataset.