Regularized Evolution for Image Classifier Architecture Search

Benchmark Model Rank Results
architecture-search-on-cifar-10-imageAmoebaNet-B + c/o#8Percentage error: 2.13Params: 34.9M
image-classification-on-imagenetAmoebaNet-A#361Top 1 Accuracy: 83.9%Number of params: 469MGFLOPs: 208
neural-architecture-search-on-nas-bench-201REA#20Accuracy (Test): 45.54Search time (s): 12000
neural-architecture-search-on-nats-benchRE (Real et al., 2019)#2Test Accuracy: 44.76
neural-architecture-search-on-nats-bench-1RE (Real et al., 2019)#1Test Accuracy: 94.13
neural-architecture-search-on-nats-bench-2RE (Real et al., 2019)#2Test Accuracy: 71.40