# Pi-NAS: Improving Neural Architecture Search by Reducing Supernet Training Consistency Shift

arXiv: https://arxiv.org/abs/2108.09671 · HTML: https://mlsanity.com/paper/2108.09671/

## Key facts

- Cited by 14 papers in the map (data through 14 Oct 2026).
- Builds on 52 of its 67 references that are in the map.
- First posted on arXiv 22 Aug 2021.

## Builds on (most cited first, 30 of 52)

- [Deep Residual Learning for Image Recognition](https://mlsanity.com/paper/1512.03385/) (2015)
- [A Simple Framework for Contrastive Learning of Visual Representations](https://mlsanity.com/paper/2002.05709/) (2020)
- [Representation Learning with Contrastive Predictive Coding](https://mlsanity.com/paper/1807.03748/) (2018)
- [Momentum Contrast for Unsupervised Visual Representation Learning](https://mlsanity.com/paper/1911.05722/) (2019)
- [Mask R-CNN](https://mlsanity.com/paper/1703.06870/) (2017)
- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [Squeeze-and-Excitation Networks](https://mlsanity.com/paper/1709.01507/) (2017)
- [Bootstrap your own latent: A new approach to self-supervised Learning](https://mlsanity.com/paper/2006.07733/) (2020)
- [Improved Baselines with Momentum Contrastive Learning](https://mlsanity.com/paper/2003.04297/) (2020)
- [Rethinking Atrous Convolution for Semantic Image Segmentation](https://mlsanity.com/paper/1706.05587/) (2017)
- [Unsupervised Feature Learning via Non-Parametric Instance-level Discrimination](https://mlsanity.com/paper/1805.01978/) (2018)
- [DARTS: Differentiable Architecture Search](https://mlsanity.com/paper/1806.09055/) (2018)
- [Neural Architecture Search with Reinforcement Learning](https://mlsanity.com/paper/1611.01578/) (2016)
- [MixMatch: A Holistic Approach to Semi-Supervised Learning](https://mlsanity.com/paper/1905.02249/) (2019)
- [Self-training with Noisy Student improves ImageNet classification](https://mlsanity.com/paper/1911.04252/) (2019)
- [Contrastive Multiview Coding](https://mlsanity.com/paper/1906.05849/) (2019)
- [Learning deep representations by mutual information estimation and maximization](https://mlsanity.com/paper/1808.06670/) (2018)
- [Unsupervised Data Augmentation for Consistency Training](https://mlsanity.com/paper/1904.12848/) (2019)
- [Temporal Ensembling for Semi-Supervised Learning](https://mlsanity.com/paper/1610.02242/) (2016)
- [MnasNet: Platform-Aware Neural Architecture Search for Mobile](https://mlsanity.com/paper/1807.11626/) (2018)
- [AugMix: A Simple Data Processing Method to Improve Robustness and Uncertainty](https://mlsanity.com/paper/1912.02781/) (2019)
- [ProxylessNAS: Direct Neural Architecture Search on Target Task and Hardware](https://mlsanity.com/paper/1812.00332/) (2018)
- [FBNet: Hardware-Aware Efficient ConvNet Design via Differentiable Neural Architecture Search](https://mlsanity.com/paper/1812.03443/) (2018)
- [ResNeSt: Split-Attention Networks](https://mlsanity.com/paper/2004.08955/) (2020)
- [Regularization With Stochastic Transformations and Perturbations for Deep Semi-Supervised Learning](https://mlsanity.com/paper/1606.04586/) (2016)
- [NAS-Bench-201: Extending the Scope of Reproducible Neural Architecture Search](https://mlsanity.com/paper/2001.00326/) (2020)
- [Single Path One-Shot Neural Architecture Search with Uniform Sampling](https://mlsanity.com/paper/1904.00420/) (2019)
- [Selective Kernel Networks](https://mlsanity.com/paper/1903.06586/) (2019)
- [Large-Scale Evolution of Image Classifiers](https://mlsanity.com/paper/1703.01041/) (2017)
- [S4L: Self-Supervised Semi-Supervised Learning](https://mlsanity.com/paper/1905.03670/) (2019)

## Built on by (most cited first, 14 of 14)

- [SparseNeRF: Distilling Depth Ranking for Few-shot Novel View Synthesis](https://mlsanity.com/paper/2303.16196/) (2023)
- [AutoDiffusion: Training-Free Optimization of Time Steps and Architectures for Automated Diffusion Model Acceleration](https://mlsanity.com/paper/2309.10438/) (2023)
- [Automated Progressive Learning for Efficient Training of Vision Transformers](https://mlsanity.com/paper/2203.14509/) (2022)
- [FrugalNeRF: Fast Convergence for Extreme Few-shot Novel View Synthesis without Learned Priors](https://mlsanity.com/paper/2410.16271/) (2024)
- [NAS-LID: Efficient Neural Architecture Search with Local Intrinsic Dimension](https://mlsanity.com/paper/2211.12759/) (2022)
- [Efficient Training of Large Vision Models via Advanced Automated Progressive Learning](https://mlsanity.com/paper/2410.00350/) (2024)
- [DNA Family: Boosting Weight-Sharing NAS with Block-Wise Supervisions](https://mlsanity.com/paper/2403.01326/) (2024)
- [Subnet-Aware Dynamic Supernet Training for Neural Architecture Search](https://mlsanity.com/paper/2503.10740/) (2025)
- [An Encode-then-Decompose Approach to Unsupervised Time Series Anomaly Detection on Contaminated Training Data--Extended Version](https://mlsanity.com/paper/2510.18998/) (2025)
- [Improve Ranking Correlation of Super-net through Training Scheme from One-shot NAS to Few-shot NAS](https://mlsanity.com/paper/2206.05896/) (2022)
- [DDS-NAS: Dynamic Data Selection within Neural Architecture Search via On-line Hard Example Mining applied to Image Classification](https://mlsanity.com/paper/2506.14667/) (2025)
- [Efficient Training for Human Video Generation with Entropy-Guided Prioritized Progressive Learning](https://mlsanity.com/paper/2511.21136/) (2025)
- [Which Layer Causes Distribution Deviation? Entropy-Guided Adaptive Pruning for Diffusion and Flow Models](https://mlsanity.com/paper/2511.21122/) (2025)
- [TAS-LoRA: Transformer Architecture Search with Mixture-of-LoRA Experts](https://mlsanity.com/paper/2605.07256/) (2026)

## Abstract (arXiv)

Recently proposed neural architecture search (NAS) methods co-train billions of architectures in a supernet and estimate their potential accuracy using the network weights detached from the supernet. However, the ranking correlation between the architectures' predicted accuracy and their actual capability is incorrect, which causes the existing NAS methods' dilemma. We attribute this ranking correlation problem to the supernet training consistency shift, including feature shift and parameter shift. Feature shift is identified as dynamic input distributions of a hidden layer due to random path sampling. The input distribution dynamic affects the loss descent and finally affects architecture ranking. Parameter shift is identified as contradictory parameter updates for a shared layer lay in different paths in different training steps. The rapidly-changing parameter could not preserve architecture ranking. We address these two shifts simultaneously using a nontrivial supernet-Pi model, called Pi-NAS. Specifically, we employ a supernet-Pi model that contains cross-path learning to reduce the feature consistency shift between different paths. Meanwhile, we adopt a novel nontrivial mean teacher containing negative samples to overcome parameter shift and model collision. Furthermore, our Pi-NAS runs in an unsupervised manner, which can search for more transferable architectures. Extensive experiments on ImageNet and a wide range of downstream tasks (e.g., COCO 2017, ADE20K, and Cityscapes) demonstrate the effectiveness and universality of our Pi-NAS compared to supervised NAS. See Codes: https://github.com/Ernie1/Pi-NAS.