Source-Free Adaptation to Measurement Shift via Bottom-Up Feature Restoration
arXiv:2107.05446
Abstract
Source-free domain adaptation (SFDA) aims to adapt a model trained on labelled data in a source domain to unlabelled data in a target domain without access to the source-domain data during adaptation. Existing methods for SFDA leverage entropy-minimization techniques which: (i) apply only to classification; (ii) destroy model calibration; and (iii) rely on the source model achieving a good level of feature-space class-separation in the target domain. We address these issues for a particularly pervasive type of domain shift called measurement shift which can be resolved by restoring the source features rather than extracting new ones. In particular, we propose Feature Restoration (FR) wherein we: (i) store a lightweight and flexible approximation of the feature distribution under the source data; and (ii) adapt the feature-extractor such that the approximate feature distribution under the target data realigns with that saved on the source. We additionally propose a bottom-up training scheme which boosts performance, which we call Bottom-Up Feature Restoration (BUFR). On real and synthetic data, we demonstrate that BUFR outperforms existing SFDA methods in terms of accuracy, calibration, and data efficiency, while being less reliant on the performance of the source model in the target domain.
arXiv commentICLR 2022 (Spotlight)
References in corpus (24)
- How transferable are features in deep neural networks?
- Learning Transferable Features with Deep Adaptation Networks
- Shortcut Learning in Deep Neural Networks
- On Calibration of Modern Neural Networks
- Confounding variables can degrade generalization performance of radiological deep learning models
- Conditional Adversarial Domain Adaptation
- ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustness
- Quantifying the effects of data augmentation and stain color normalization in convolutional neural networks for computational pathology
- Model Adaptation: Unsupervised Domain Adaptation without Source Data
- EMNIST: an extension of MNIST to handwritten letters
- A DIRT-T Approach to Unsupervised Domain Adaptation
- Generalisation in humans and deep neural networks
- Revisiting Batch Normalization For Practical Domain Adaptation
- Invariant Risk Minimization
- Tent: Fully Test-time Adaptation by Entropy Minimization
- Exploring the Landscape of Spatial Robustness
- Unsupervised Domain Adaptation through Self-Supervision
- Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Coming
- Deep Neural Decision Trees
- Evaluating Prediction-Time Batch Normalization for Robustness under Covariate Shift
- Towards Accurate Model Selection in Deep Unsupervised Domain Adaptation
- Effects of Degradations on Deep Neural Network Architectures
- Source-free Domain Adaptation via Distributional Alignment by Matching Batch Normalization Statistics
- A Flexible Selection Scheme for Minimum-Effort Transfer Learning