# Rethinking Importance Weighting for Deep Learning under Distribution Shift

arXiv: https://arxiv.org/abs/2006.04662 · HTML: https://mlsanity.com/paper/2006.04662/

## Key facts

- Cited by 93 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 8 Jun 2020.

## Built on by (most cited first, 30 of 93)

- [NICO++: Towards Better Benchmarking for Domain Generalization](https://mlsanity.com/paper/2204.08040/) (2022)
- [Mandoline: Model Evaluation under Distribution Shift](https://mlsanity.com/paper/2107.00643/) (2021)
- [Model Agnostic Sample Reweighting for Out-of-Distribution Learning](https://mlsanity.com/paper/2301.09819/) (2023)
- [Understanding the role of importance weighting for deep learning](https://mlsanity.com/paper/2103.15209/) (2021)
- [Unleashing the Power of Graph Data Augmentation on Covariate Distribution Shift](https://mlsanity.com/paper/2211.02843/) (2022)
- [Performative Prediction with Neural Networks](https://mlsanity.com/paper/2304.06879/) (2023)
- [Meta Discovery: Learning to Discover Novel Classes given Very Limited Data](https://mlsanity.com/paper/2102.04002/) (2021)
- [Understanding Why Generalized Reweighting Does Not Improve Over ERM](https://mlsanity.com/paper/2201.12293/) (2022)
- [Learning Bounds for Open-Set Learning](https://mlsanity.com/paper/2106.15792/) (2021)
- [Operationalizing Machine Learning: An Interview Study](https://mlsanity.com/paper/2209.09125/) (2022)
- [Learning to Re-weight Examples with Optimal Transport for Imbalanced Classification](https://mlsanity.com/paper/2208.02951/) (2022)
- [DGMamba: Domain Generalization via Generalized State Space Model](https://mlsanity.com/paper/2404.07794/) (2024)
- [A Theoretical Analysis on Independence-driven Importance Weighting for Covariate-shift Generalization](https://mlsanity.com/paper/2111.02355/) (2021)
- [TOHAN: A One-step Approach towards Few-shot Hypothesis Adaptation](https://mlsanity.com/paper/2106.06326/) (2021)
- [Rethinking Domain Generalization: Discriminability and Generalizability](https://mlsanity.com/paper/2309.16483/) (2023)
- [Dynamic Loss-Based Sample Reweighting for Improved Large Language Model Pretraining](https://mlsanity.com/paper/2502.06733/) (2025)
- [Adapting to Continuous Covariate Shift via Online Density Ratio Estimation](https://mlsanity.com/paper/2302.02552/) (2023)
- [A Short Survey on Importance Weighting for Machine Learning](https://mlsanity.com/paper/2403.10175/) (2024)
- [Self-Adaptive Forecasting for Improved Deep Learning on Non-Stationary Time-Series](https://mlsanity.com/paper/2202.02403/) (2022)
- [When is Importance Weighting Correction Needed for Covariate Shift Adaptation?](https://mlsanity.com/paper/2303.04020/) (2023)
- [Binary Classification from Multiple Unlabeled Datasets via Surrogate Set Classification](https://mlsanity.com/paper/2102.00678/) (2021)
- [Importance Weighting Can Help Large Language Models Self-Improve](https://mlsanity.com/paper/2408.09849/) (2024)
- [Rethinking Importance Weighting for Transfer Learning](https://mlsanity.com/paper/2112.10157/) (2021)
- [LOPS: Learning Order Inspired Pseudo-Label Selection for Weakly Supervised Text Classification](https://mlsanity.com/paper/2205.12528/) (2022)
- [L2B: Learning to Bootstrap Robust Models for Combating Label Noise](https://mlsanity.com/paper/2202.04291/) (2022)
- [Generalizing Importance Weighting to A Universal Solver for Distribution Shift Problems](https://mlsanity.com/paper/2305.14690/) (2023)
- [RegExplainer: Generating Explanations for Graph Neural Networks in Regression Tasks](https://mlsanity.com/paper/2307.07840/) (2023)
- [Foster Adaptivity and Balance in Learning with Noisy Labels](https://mlsanity.com/paper/2407.02778/) (2024)
- [Distribution Shift Matters for Knowledge Distillation with Webly Collected Images](https://mlsanity.com/paper/2307.11469/) (2023)
- [When to retrain a machine learning model](https://mlsanity.com/paper/2505.14903/) (2025)

## Abstract (arXiv)

Under distribution shift (DS) where the training data distribution differs from the test one, a powerful technique is importance weighting (IW) which handles DS in two separate steps: weight estimation (WE) estimates the test-over-training density ratio and weighted classification (WC) trains the classifier from weighted training data. However, IW cannot work well on complex data, since WE is incompatible with deep learning. In this paper, we rethink IW and theoretically show it suffers from a circular dependency: we need not only WE for WC, but also WC for WE where a trained deep classifier is used as the feature extractor (FE). To cut off the dependency, we try to pretrain FE from unweighted training data, which leads to biased FE. To overcome the bias, we propose an end-to-end solution dynamic IW that iterates between WE and WC and combines them in a seamless manner, and hence our WE can also enjoy deep networks and stochastic optimizers indirectly. Experiments with two representative types of DS on three popular datasets show that our dynamic IW compares favorably with state-of-the-art methods.