AlphaNet: Improved Training of Supernets with Alpha-Divergence

Benchmark Model Rank Results
image-classification-on-imagenetAlphaNet-A6#650Top 1 Accuracy: 80.8%GFLOPs: 0.709
image-classification-on-imagenetAlphaNet-A5#676Top 1 Accuracy: 80.3%GFLOPs: 0.491
image-classification-on-imagenetAlphaNet-A4#687Top 1 Accuracy: 80.0%GFLOPs: 0.444
image-classification-on-imagenetAlphaNet-A3#717Top 1 Accuracy: 79.4%GFLOPs: 0.357
image-classification-on-imagenetAlphaNet-A2#734Top 1 Accuracy: 79.1%GFLOPs: 0.317
image-classification-on-imagenetAlphaNet-A1#757Top 1 Accuracy: 78.9%GFLOPs: 0.279
image-classification-on-imagenetAlphaNet-A0#812Top 1 Accuracy: 77.8%GFLOPs: 0.203
neural-architecture-search-on-imagenetAlphaNet-A6#10Top-1 Error Rate: 19.2Accuracy: 80.8FLOPs: 709M
neural-architecture-search-on-imagenetAlphaNet-A5 (base)#12Top-1 Error Rate: 19.4Accuracy: 80.6FLOPs: 596M
neural-architecture-search-on-imagenetAlphaNet-A5 (small)#17Top-1 Error Rate: 19.7Accuracy: 80.3FLOPs: 491M
neural-architecture-search-on-imagenetAlphaNet-A4#21Top-1 Error Rate: 20.0Accuracy: 80.0FLOPs: 444M
neural-architecture-search-on-imagenetAlphaNet-A3#31Top-1 Error Rate: 20.6Accuracy: 79.4FLOPs: 357M
neural-architecture-search-on-imagenetAlphaNet-A2#32Top-1 Error Rate: 20.8Accuracy: 79.2FLOPs: 317M
neural-architecture-search-on-imagenetAlphaNet-A1#35Top-1 Error Rate: 21.0Accuracy: 79.0FLOPs: 279M
neural-architecture-search-on-imagenetAlphaNet-A0#51Top-1 Error Rate: 22.1Accuracy: 77.9FLOPs: 203M