Regularized Evolution for Image Classifier Architecture Search

5 Feb 2018  ·  Esteban Real, Alok Aggarwal, Yanping Huang, Quoc V. Le ·

The effort devoted to hand-crafting neural network image classifiers has motivated the use of architecture search to discover them automatically. Although evolutionary algorithms have been repeatedly applied to neural network topologies, the image classifiers thus discovered have remained inferior to human-crafted ones. Here, we evolve an image classifier---AmoebaNet-A---that surpasses hand-designs for the first time. To do this, we modify the tournament selection evolutionary algorithm by introducing an age property to favor the younger genotypes. Matching size, AmoebaNet-A has comparable accuracy to current state-of-the-art ImageNet models discovered with more complex architecture-search methods. Scaled to larger size, AmoebaNet-A sets a new state-of-the-art 83.9% / 96.6% top-5 ImageNet accuracy. In a controlled comparison against a well known reinforcement learning algorithm, we give evidence that evolution can obtain results faster with the same hardware, especially at the earlier stages of the search. This is relevant when fewer compute resources are available. Evolution is, thus, a simple method to effectively discover high-quality architectures.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Result Benchmark
Neural Architecture Search CIFAR-10 Image Classification AmoebaNet-B + c/o Percentage error 2.13 # 8
Params 34.9M # 18
Image Classification ImageNet AmoebaNet-A Top 1 Accuracy 83.9% # 181
Top 5 Accuracy 96.6% # 55
Number of params 469M # 30
Hardware Burden None # 1
Operations per network pass None # 1
Neural Architecture Search NAS-Bench-201, ImageNet-16-120 REA Accuracy (Test) 45.54 # 11
Accuracy (val) 45.15 # 6
Search time (s) 12000 # 11