Spartan: Differentiable Sparsity via Regularized Transportation
This addresses the need for efficient, sparse models in deep learning, particularly for resource-constrained applications, and is incremental as it builds on existing sparsity techniques with novel regularization and optimization methods.
The paper tackles the problem of training sparse neural networks with predetermined sparsity levels, achieving results such as 95% sparse ResNet-50 and 90% block sparse ViT-B/16 models with less than 1% absolute top-1 accuracy loss on ImageNet-1K compared to dense training.
We present Spartan, a method for training sparse neural network models with a predetermined level of sparsity. Spartan is based on a combination of two techniques: (1) soft top-k masking of low-magnitude parameters via a regularized optimal transportation problem and (2) dual averaging-based parameter updates with hard sparsification in the forward pass. This scheme realizes an exploration-exploitation tradeoff: early in training, the learner is able to explore various sparsity patterns, and as the soft top-k approximation is gradually sharpened over the course of training, the balance shifts towards parameter optimization with respect to a fixed sparsity mask. Spartan is sufficiently flexible to accommodate a variety of sparsity allocation policies, including both unstructured and block structured sparsity, as well as general cost-sensitive sparsity allocation mediated by linear models of per-parameter costs. On ImageNet-1K classification, Spartan yields 95% sparse ResNet-50 models and 90% block sparse ViT-B/16 models while incurring absolute top-1 accuracy losses of less than 1% compared to fully dense training.