Research map / Architectures and efficiency / Neural architecture search
FairNAS: Rethinking Evaluation Fairness of Weight Sharing Neural Architecture Search
Xiangxiang Chu, Bo Zhang, Ruijun Xu
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add strict fairness (m models per step, accumulated gradients) instead of single-path SPOS sampling (ablation) | supernet training · ImageNet search space (a), 13 models | Kendall tau: 0.6153 -> 0.9487 | Improved |
| add strict fairness instead of SPOS single-path sampling (ablation) | supernet training · NAS-Bench-201 | Kendall tau: 0.5681 -> 0.7412 | Improved |
| add strict fairness instead of expectation-fairness uniform sampling with k=6 (ablation) | supernet training · ImageNet search space (a), 13 models | Kendall tau: 0.4871 -> 0.9487 | Improved |
| add expectation fairness (SPOS) over the unfair One-Shot supernet (ablation) | supernet training · ImageNet search space (a), 13 models | Kendall tau: 0.1245 -> 0.6153 | Improved |
| scale learning rate to 1/6 for uniform expectation-fairness training (ablation) | Uniform (k=6) supernet · NAS-Bench-201 | Kendall tau: 0.3651 -> 0.4072 | Mixed |
| add strict-fairness supernet (FairNAS) instead of One-Shot or SPOS, constrained to 400M FLOPS (ablation) | searched ImageNet model · ImageNet val | top-1 accuracy (%): 74.6 -> 75.3 | Improved |
| add SE, Swish and AutoAugment to FairNAS-A | FairNAS-A · ImageNet val | top-1 accuracy (%): 75.3 -> 77.5 | Improved |
| use FairNAS instead of GDAS on NAS-Bench-201 | NAS-Bench-201 search · ImageNet16-120 test | accuracy (%): 41.71 -> 42.19 | Improved |
| use FairNAS instead of GDAS on NAS-Bench-201 | NAS-Bench-201 search · CIFAR-10 test | accuracy (%): 93.61 -> 93.23 | Hurt |
| use FairNAS backbone instead of MixNet-M in RetinaNet | RetinaNet · COCO val2017 | AP (%): 31.3 -> 32.4 | Improved |
| use FairNAS-A instead of MobileNetV2 in DeepLabv3+ | DeepLabv3+ · PASCAL VOC 2012 | mIOU: 75.32 -> 78.54 | Improved |
| use the multi-objective EA search (NSGA-II with reinforced mutation, PPO) instead of random search or MnasNet RL (ablation) | stage 2 search · ImageNet search space (a) | Pareto front quality | Improved |
Later papers that compared with this method
3 later accepted papers report FairNAS as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Boosting Order-Preserving and Transferability for Neural Architecture Search: a Joint Architecture Refined Search and Fine-tuning Approach (CVPR 2024 · better on 3 of 3 shared results)
- BN-NAS: Neural Architecture Search With Batch Normalization (ICCV 2021 · better on 3 of 3 shared results)
- Pi-NAS: Improving Neural Architecture Search by Reducing Supernet Training Consistency Shift (ICCV 2021 · better on 2 of 2 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Fair Bilevel Neural Network (FairBiNN): On Balancing fairness and accuracy via Stackelberg Equilibrium (NeurIPS 2024 · similarity 0.87)
- FairNet: Dynamic Fairness Correction without Performance Loss via Contrastive Conditional LoRA (NeurIPS 2025 · similarity 0.86)
- Some Optimizers are More Equal: Understanding the Role of Optimizers in Group Fairness (NeurIPS 2025 · similarity 0.85)
- Chasing Fairness in Graphs: A GNN Architecture Perspective (AAAI 2024 · similarity 0.85)
- FairSIN: Achieving Fairness in Graph Neural Networks through Sensitive Information Neutralization (AAAI 2024 · similarity 0.85)
- Classes Are Not Equal: An Empirical Study on Image Recognition Fairness (CVPR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
