Open-Set Domain Adaptation Under Background Distribution Shift: Challenges and A Provably Efficient Solution
cs.LG, cs.AI, cs.CV
Submitted: 2025-12-01
Updated: 2026-09-06
Comments: Project page at https://github.com/Shra1-25/CoLOR
Journal ref: Transactions on Machine Learning Research (TMLR) 2026/May ISSN: 2835-8856
Code: https://github.com/Shra1-25/CoLOR
License: http://creativecommons.org/licenses/by/4.0/
The gist: As we deploy machine learning systems in the real world, a core challenge is to maintain a model that is performant even as the data shifts.
Terminology
Abstract
As we deploy machine learning systems in the real world, a core challenge is to maintain a model that is performant even as the data shifts. Such shifts can take many forms: new classes may emerge that were absent during training, a problem known as open-set recognition, and the distribution of known categories may change. Guarantees on open-set recognition are mostly derived under the assumption that the distribution of known classes, which we call the background distribution, is fixed. In this paper we develop CoLOR, a method that is guaranteed to solve open-set recognition even in the challenging case where the background distribution shifts. We prove that the method works under benign assumptions that the novel class is separable from the non-novel classes, and provide theoretical guarantees that it outperforms a representative baseline in a simplified overparameterized setting. We develop techniques to make CoLOR scalable and robust, and perform comprehensive empirical evaluations on image and text data. The results show that CoLOR significantly outperforms existing open-set recognition methods under background shift. Moreover, we provide new insights into how factors such as the size of the novel class influences performance, an aspect that has not been extensively explored in prior work.
Sources
- Concrete Problems in AI Safety
- A Model of Inductive Bias Learning
- Anomaly Detection under Distribution Shift
- Domain Adaptation under Open Set Label Shift
- Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation
- RoBERTa: A Robustly Optimized BERT Pretraining Approach
- Learning Transferable Visual Models From Natural Language Supervision
- A survey on domain adaptation theory: learning bounds and theoretical guarantees
- Deep Hashing Network for Unsupervised Domain Adaptation
- Malign Overfitting: Interpolation Can Provably Preclude Invariance
- Model-free Test Time Adaptation for Out-Of-Distribution Detection
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks