Neural Architecture Search On Imagenet
Metrics
Accuracy
FLOPs
Params
Top-1 Error Rate
Results
Performance results of various models on this benchmark
Comparison Table
| Model Name | Accuracy | FLOPs | Params | Top-1 Error Rate |
|---|---|---|---|---|
| blockwisely-supervised-neural-architecture | 78.4 | 611M | 6.4M | 21.6 |
| b-darts-beta-decay-regularization-for | - | - | - | 23.9 |
| hardcore-nas-hard-constrained-differentiable | 75.9 | - | - | 24.1 |
| darts-differentiable-architecture-search | 73.3 | - | 4.9 | 26.7 |
| m-darts-model-uncertainty-aware | 78.76 | - | 602M | 21.24 |
| alphanet-improved-training-of-supernet-with | 79.2 | 317M | - | 20.8 |
| neural-architecture-transfer | 80.5 | - | 9.1M | 19.5 |
| evolving-neural-architecture-using-one-shot | 74.9 | - | 4.9 | 25.1 |
| single-path-one-shot-neural-architecture | 75.1 | - | - | - |
| densely-connected-search-space-for-more | - | 479M | - | 23.9 |
| hardcore-nas-hard-constrained-differentiable | 80.1 | - | - | 19.9 |
| isynet-convolutional-neural-networks-design | - | - | - | 22.7 |
| hardcore-nas-hard-constrained-differentiable | 79.5 | - | - | 20.5 |
| idarts-improving-darts-by-node-normalization | - | - | 5.1M | 24.7 |
| sharpdarts-faster-and-more-accurate | 74.1 | - | 4.9M | 25.1 |
| deepmad-mathematical-architecture-design-for | 83.9 | 8.7G | 50M | 16.1 |
| alphanet-improved-training-of-supernet-with | 79.0 | 279M | - | 21.0 |
| alphanet-improved-training-of-supernet-with | 80.8 | 709M | - | 19.2 |
| evolving-neural-architecture-using-one-shot | 75.6 | - | 5.3 | 24.4 |
| fairnas-rethinking-evaluation-fairness-of | 75.1 | - | 4.5M | 24.9 |
| single-path-one-shot-neural-architecture | 74.7 | - | - | - |
| eeea-net-an-early-exit-evolutionary-neural | 76.2 | - | - | 23.8 |
| attentivenas-improving-neural-architecture | 80.1 | - | - | 19.9 |
| attentivenas-improving-neural-architecture | 77.3 | - | - | 22.7 |
| alphanet-improved-training-of-supernet-with | 77.9 | 203M | - | 22.1 |
| uninet-unified-architecture-search-with-1 | - | 555M | 11.5M | 19.2 |
| nasvit-neural-architecture-search-for | 81.4 | 591M | - | 18.6 |
| blockwisely-supervised-neural-architecture | 77.8 | 466M | 5.3M | 22.2 |
| layernas-neural-architecture-search-in | - | - | 3.7M | 31 |
| how-does-topology-of-neural-architectures | 73.3 | 393M | 3.7M | 26.7 |
| isynet-convolutional-neural-networks-design | - | - | - | 23.16 |
| alphanet-improved-training-of-supernet-with | 79.4 | 357M | - | 20.6 |
| alphanet-improved-training-of-supernet-with | 80.3 | 491M | - | 19.7 |
| hardcore-nas-hard-constrained-differentiable | 77.1 | - | - | 22.9 |
| hardcore-nas-hard-constrained-differentiable | 78.8 | - | - | 21.2 |
| fbnetv5-neural-architecture-search-for | - | 726M | - | 18.2 |
| nasvit-neural-architecture-search-for | 78.2 | 208M | - | 21.8 |
| sharpdarts-faster-and-more-accurate | 76.0 | - | 8.3M | 24.0 |
| du-darts-decreasing-the-uncertainty-of | 75.9 | - | 5.3M | 24.1 |
| alphanet-improved-training-of-supernet-with | 80.6 | 596M | - | 19.4 |
| nasvit-neural-architecture-search-for | 79.7 | 309M | - | 20.3 |
| once-for-all-train-one-network-and-specialize | 76.9 | - | - | 23.1 |
| hardcore-nas-hard-constrained-differentiable | 77.9 | - | - | 22.1 |
| neural-architecture-transfer | 79.9 | - | 9.1M | 20.1 |
| isynet-convolutional-neural-networks-design | - | - | - | 21.45 |
| isynet-convolutional-neural-networks-design | - | - | - | 24.55 |
| greedynas-towards-fast-one-shot-nas-with | 76.8 | - | 5.2M | 23.2 |
| bignas-scaling-up-neural-architecture-search | 78.9 | - | 5.5M | 21.1 |
| sgas-sequential-greedy-architecture-search | 75.9 | - | 5.4M | 24.1 |
| nsganetv2-evolutionary-multi-objective | 80.4 | - | 8.7M | 19.6 |
| fbnetv5-neural-architecture-search-for | 81.7 | 685M | - | 18.3 |
| fbnetv3-joint-architecture-recipe-search | 79.6 | - | - | 20.4 |
| geometry-aware-gradient-algorithms-for-neural | - | - | 5.6 | 24 |
| gpunet-searching-the-deployable-convolution | - | 3.66G | 10.6M | 17.5 |
| single-path-one-shot-neural-architecture | 75.3 | - | - | - |
| fbnetv2-differentiable-neural-architecture | 77.2 | - | - | 22.8 |
| blockwisely-supervised-neural-architecture | 77.5 | 406M | 4.9M | 22.5 |
| fairnas-rethinking-evaluation-fairness-of | 75.34 | - | 4.6M | 24.7 |
| bignas-scaling-up-neural-architecture-search | 76.5 | - | 4.5M | 23.5 |
| neural-architecture-search-for-lightweight | 77.7 | - | - | 22.3 |
| layernas-neural-architecture-search-in | - | - | 9.7M | 21.4 |
| neural-architecture-search-for-lightweight | 76.5 | - | - | 23.5 |
| atomnas-fine-grained-end-to-end-neural-1 | 77.2 | - | 5.5M | 22.8 |
| fairnas-rethinking-evaluation-fairness-of | 74.69 | - | 4.4M | 25.4 |
| muxconv-information-multiplexing-in | 76.6 | - | 4.0M | 23.4 |
| memnas-memory-efficient-neural-architecture | 74.1 | - | - | 25.9 |
| fbnetv2-differentiable-neural-architecture | 73.2 | - | - | 26.8 |
| optimizing-neural-architecture-search-using | 70.49 | - | - | 29.51 |
| sedona-search-for-decoupled-neural-networks | - | - | - | 21.09 |
| evolving-neural-architecture-using-one-shot | 75.6 | - | 5.1 | 24.4 |
| isynet-convolutional-neural-networks-design | - | - | - | 22.15 |
| layernas-neural-architecture-search-in | - | - | 5.2M | 22.9 |
| neural-architecture-transfer | 77.5 | - | 6.0M | 22.5 |
| scarletnas-bridging-the-gap-between | 76.3 | - | 6.5M | 23.7 |
| fbnetv2-differentiable-neural-architecture | 68.3 | - | - | 31.7 |
| Model 76 | 75.43 | - | - | 24.57 |
| zen-nas-a-zero-shot-nas-for-high-performance | 77.8 | 1.7G | 30.1 | 22.2 |
| alphax-exploring-neural-architectures-with-1 | 75.5 | - | 5.4M | 24.5 |
| how-does-topology-of-neural-architectures | 72.9 | 200M | 2.3M | 27.1 |
| muxconv-information-multiplexing-in | 75.3 | - | 3.4M | 24.7 |
| exploring-randomly-wired-neural-networks-for | - | 583M | 5.6M | 25.3 |
| attentivenas-improving-neural-architecture | 79.8 | - | - | 20.2 |
| atomnas-fine-grained-end-to-end-neural-1 | 77.6 | - | 5.9M | 22.4 |
| fbnetv5-neural-architecture-search-for | 77.2 | 215M | - | 22.8 |
| pc-darts-partial-channel-connections-for | 75.8 | - | 5.3M | 24.2 |
| layernas-neural-architecture-search-in | - | - | 5.1M | 24.4 |
| isynet-convolutional-neural-networks-design | - | - | - | 19.84 |
| nsganetv2-evolutionary-multi-objective | 79.1 | - | 8.0M | 20.9 |
| zen-nas-a-zero-shot-nas-for-high-performance | 83.6 | 22G | 180M | 16.4 |
| gpunet-searching-the-deployable-convolution | - | 720M | 6.2M | 20.3 |
| gpunet-searching-the-deployable-convolution | - | 15.6G | 19M | 16.4 |
| a-generic-graph-based-neural-architecture | 75.9 | - | 5.6M | 24.1 |
| fbnetv3-joint-architecture-recipe-search | 78.0 | - | - | 22.0 |
| scarletnas-bridging-the-gap-between | 75.6 | - | 6.0M | 24.4 |
| nasvit-neural-architecture-search-for | 81.8 | 757M | - | 18.2 |
| proxylessnas-direct-neural-architecture | 75.1 | - | 5.1M | 24.9 |
| fbnetv3-joint-architecture-recipe-search | 82.3 | - | - | 17.7 |
| fair-darts-eliminating-unfair-advantages-in | - | - | - | 22.8 |
| atomnas-fine-grained-end-to-end-neural-1 | 76.3 | - | 4.7M | 23.7 |
| hardcore-nas-hard-constrained-differentiable | 78.3 | - | - | 21.7 |
| bossnas-exploring-hybrid-cnn-transformers | 82.2 | - | - | 17.8 |
| attentivenas-improving-neural-architecture | 78.4 | - | - | 21.6 |
| hardcore-nas-hard-constrained-differentiable | 77.4 | - | - | 22.6 |
| eeea-net-an-early-exit-evolutionary-neural | 74.3 | - | - | 25.7 |
| hardcore-nas-hard-constrained-differentiable | 76.5 | - | - | 23.5 |
| greedynas-towards-fast-one-shot-nas-with | 76.2 | - | 4.7M | 23.8 |
| muxconv-information-multiplexing-in | 71.6 | - | 2.4M | 28.4 |
| nasvit-neural-architecture-search-for | 81.0 | 528M | - | 19.0 |
| drnas-dirichlet-neural-architecture-search | - | - | 5.7M | 23.7 |
| muxconv-information-multiplexing-in | 66.7 | - | 1.8M | 33.3 |
| nsganetv2-evolutionary-multi-objective | 77.4 | - | 6.1M | 22.6 |
| nsganetv2-evolutionary-multi-objective | 78.3 | - | 7.7M | 21.7 |
| attentivenas-improving-neural-architecture | 78.8 | - | - | 21.2 |
| attentivenas-improving-neural-architecture | 79.1 | - | - | 20.9 |
| neural-architecture-transfer | 78.6 | - | 7.7M | 21.4 |
| idarts-improving-darts-by-node-normalization | - | - | 5.1M | 25.2 |
| hardcore-nas-hard-constrained-differentiable | 78.1 | - | - | 21.9 |
| scarletnas-bridging-the-gap-between | 76.9 | - | 6.7M | 23.1 |
| hardcore-nas-hard-constrained-differentiable | 78.9 | - | - | 21.1 |
| greedynas-towards-fast-one-shot-nas-with | 77.1 | - | 6.5M | 22.9 |
| blockwisely-supervised-neural-architecture | 77.1 | 348M | 4.2M | 22.9 |
| bignas-scaling-up-neural-architecture-search | 79.5 | - | 6.4M | 20.5 |
| isynet-convolutional-neural-networks-design | - | - | - | 20.59 |
| fbnetv2-differentiable-neural-architecture | 76.0 | - | - | 24.0 |
| neural-architecture-search-using-stable-rank | 76.1 | - | 5.6M | 23.9 |
| nasvit-neural-architecture-search-for | 80.5 | 421M | - | 19.5 |
| shapley-nas-discovering-operation-1 | - | - | 5.4M | 23.9 |
| fbnetv3-joint-architecture-recipe-search | 80.4 | - | - | 19.6 |
| sedona-search-for-decoupled-neural-networks | - | - | - | 20.2 |
| semi-supervised-neural-architecture-search | 76.5 | - | - | 23.5 |
| noisy-differentiable-architecture-search | 77.9 | - | 5.5M | 22.1 |
| progressive-neural-architecture-search | - | - | 5.1 | - |
| Model 133 | 74.0 | - | 5.3 | 26.0 |
| alphanet-improved-training-of-supernet-with | 80.0 | 444M | - | 20.0 |
| neural-architecture-search-with-gbdt | 76.5 | - | 6.4M | 23.5 |