Multiscale Deep Equilibrium Models
This addresses the problem of scaling implicit networks for large-scale vision tasks, representing a significant advancement in the field.
The authors tackled large-scale hierarchical pattern recognition by proposing multiscale deep equilibrium models (MDEQs), which solve for equilibrium points across multiple feature resolutions with O(1) memory consumption, enabling a single model to match or exceed performance on ImageNet classification and Cityscapes segmentation compared to recent competitive models.
We propose a new class of implicit networks, the multiscale deep equilibrium model (MDEQ), suited to large-scale and highly hierarchical pattern recognition domains. An MDEQ directly solves for and backpropagates through the equilibrium points of multiple feature resolutions simultaneously, using implicit differentiation to avoid storing intermediate states (and thus requiring only $O(1)$ memory consumption). These simultaneously-learned multi-resolution features allow us to train a single model on a diverse set of tasks and loss functions, such as using a single MDEQ to perform both image classification and semantic segmentation. We illustrate the effectiveness of this approach on two large-scale vision tasks: ImageNet classification and semantic segmentation on high-resolution images from the Cityscapes dataset. In both settings, MDEQs are able to match or exceed the performance of recent competitive computer vision models: the first time such performance and scale have been achieved by an implicit deep learning approach. The code and pre-trained models are at https://github.com/locuslab/mdeq .