Towards Data-Free Domain Generalization

Abstract

In this work, we investigate the unexplored intersection of domain generalization (DG) and data-free learning. In particular, we address the question: How can knowledge contained in models trained on different source domains be merged into a single model that generalizes well to unseen target domains, in the absence of source and target domain data? Machine learning models that can cope with domain shift are essential for real-world scenarios with often changing data distributions. Prior DG methods typically rely on using source domain data, making them unsuitable for private decentralized data. We define the novel problem of Data-Free Domain Generalization (DFDG), a practical setting where models trained on the source domains separately are available instead of the original datasets, and investigate how to effectively solve the domain generalization problem in that case. We propose DEKAN, an approach that extracts and fuses domain-specific knowledge from the available teacher models into a student model robust to domain shift. Our empirical evaluation demonstrates the effectiveness of our method which achieves first state-of-the-art results in DFDG by significantly outperforming data-free knowledge distillation and ensemble baselines.

arXiv commentAccepted at NeurIPS 2021 (DistShift Workshop) and ACML 2022

References in corpus (32)

• Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift
• Distilling the Knowledge in a Neural Network
• Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks
• mixup: Beyond Empirical Risk Minimization
• Improved Regularization of Convolutional Neural Networks with Cutout
• Deep Domain Confusion: Maximizing for Domain Invariance
• Deep Visual Domain Adaptation: A Survey
• Domain Generalization via Invariant Feature Representation
• Model Adaptation: Unsupervised Domain Adaptation without Source Data
• Domain Generalization via Model-Agnostic Learning of Semantic Features
• Generalisation in humans and deep neural networks
• Revisiting Batch Normalization For Practical Domain Adaptation
• Generalizing to Unseen Domains via Adversarial Data Augmentation
• Generalizing Across Domains via Cross-Gradient Training
• Certifying Some Distributional Robustness with Principled Adversarial Training
• Tent: Fully Test-time Adaptation by Entropy Minimization
• Model compression via distillation and quantization
• Data-Free Knowledge Distillation for Deep Neural Networks
• Apprentice: Using Knowledge Distillation Techniques To Improve Low-Precision Network Accuracy
• Improve Unsupervised Domain Adaptation with Mixup Training
• Heterogeneous Domain Generalization via Domain Mixup
• Zero-Shot Knowledge Distillation in Deep Networks
• In Search of Lost Domain Generalization
• Domain Generalization using Causal Matching
• Self-supervised Knowledge Distillation for Few-shot Learning
• Frustratingly Simple Domain Generalization via Image Stylization
• Domain Generalization with MixStyle
• Large-Scale Generative Data-Free Distillation
• SAND-mask: An Enhanced Gradient Masking Strategy for the Discovery of Invariances in Domain Generalization
• Gradient Matching for Domain Generalization
• Domain Adaptation Through Task Distillation
• Source-Free Adaptation to Measurement Shift via Bottom-Up Feature Restoration