sharpDARTS: Faster and More Accurate Differentiable Architecture Search

23 Mar 2019  ·  Andrew Hundt, Varun Jain, Gregory D. Hager ·

Neural Architecture Search (NAS) has been a source of dramatic improvements in neural network design, with recent results meeting or exceeding the performance of hand-tuned architectures. However, our understanding of how to represent the search space for neural net architectures and how to search that space efficiently are both still in their infancy. We have performed an in-depth analysis to identify limitations in a widely used search space and a recent architecture search method, Differentiable Architecture Search (DARTS). These findings led us to introduce novel network blocks with a more general, balanced, and consistent design; a better-optimized Cosine Power Annealing learning rate schedule; and other improvements. Our resulting sharpDARTS search is 50% faster with a 20-30% relative improvement in final model error on CIFAR-10 when compared to DARTS. Our best single model run has 1.93% (1.98+/-0.07) validation error on CIFAR-10 and 5.5% error (5.8+/-0.3) on the recently released CIFAR-10.1 test set. To our knowledge, both are state of the art for models of similar size. This model also generalizes competitively to ImageNet at 25.1% top-1 (7.8% top-5) error. We found improvements for existing search spaces but does DARTS generalize to new domains? We propose Differentiable Hyperparameter Grid Search and the HyperCuboid search space, which are representations designed to leverage DARTS for more general parameter optimization. Here we find that DARTS fails to generalize when compared against a human's one shot choice of models. We look back to the DARTS and sharpDARTS search spaces to understand why, and an ablation study reveals an unusual generalization gap. We finally propose Max-W regularization to solve this problem, which proves significantly better than the handmade design. Code will be made available.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Neural Architecture Search CIFAR-10 SharpSepConvDARTS Top-1 Error Rate 1.98% # 6
Search Time (GPU days) 0.8 # 17
Parameters 3.6M # 23
FLOPS 579M # 38
Neural Architecture Search CIFAR-10 sharpDARTS Top-1 Error Rate 2.29% # 10
Search Time (GPU days) 1.8 # 24
Parameters 1.98M # 17
FLOPS 357M # 33
Neural Architecture Search CIFAR-10 Image Classification SharpSepConvDARTS Percentage error 1.98 # 4
Params 3.6M # 8
FLOPS 579M # 19
Neural Architecture Search ImageNet sharpDARTS Top-1 Error Rate 24.0 # 95
Accuracy 76.0 # 75
Params 8.3M # 11
MACs 950M # 135
Neural Architecture Search ImageNet SharpSepConvDARTS Top-1 Error Rate 25.1 # 115
Accuracy 74.1 # 97
Params 4.9M # 45
MACs 573M # 121

Methods