iDARTS: Improving DARTS by Node Normalization and Decorrelation Discretization

25 Aug 2021  ·  Huiqun Wang, Ruijie Yang, Di Huang, Yunhong Wang ·

Differentiable ARchiTecture Search (DARTS) uses a continuous relaxation of network representation and dramatically accelerates Neural Architecture Search (NAS) by almost thousands of times in GPU-day. However, the searching process of DARTS is unstable, which suffers severe degradation when training epochs become large, thus limiting its application. In this paper, we claim that this degradation issue is caused by the imbalanced norms between different nodes and the highly correlated outputs from various operations. We then propose an improved version of DARTS, namely iDARTS, to deal with the two problems. In the training phase, it introduces node normalization to maintain the norm balance. In the discretization phase, the continuous architecture is approximated based on the similarity between the outputs of the node and the decorrelated operations rather than the values of the architecture parameters. Extensive evaluation is conducted on CIFAR-10 and ImageNet, and the error rates of 2.25\% and 24.7\% are reported within 0.2 and 1.9 GPU-day for architecture search respectively, which shows its effectiveness. Additional analysis also reveals that iDARTS has the advantage in robustness and generalization over other DARTS-based counterparts.

PDF Abstract
No code implementations yet. Submit your code now

Results from the Paper


Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Neural Architecture Search CIFAR-10 iDARTS +ME Top-1 Error Rate 2.25% # 9
Search Time (GPU days) 0.4 # 13
Parameters 3.6M # 23
Neural Architecture Search ImageNet iDARTS (ImageNet) Top-1 Error Rate 24.7 # 110
Params 5.1M # 41
MACs 568M # 120
Neural Architecture Search ImageNet iDARTS (CIFAR-10) Top-1 Error Rate 25.2 # 117
Params 5.1M # 41
MACs 578M # 123

Methods