Tunable Domain Adaptation Using Unfolding

2026-03-27

AI Summary

This work addresses the limited generalization of machine learning models under distribution shifts—such as varying noise levels—by proposing two tunable domain adaptation methods based on interpretable unfolded networks. The first, P-TDA, dynamically adjusts the model using known domain parameters, while the second, DD-TDA, infers domain information directly from the input to enable adaptation. This study is the first to integrate unfolded networks with domain-dependent tunable parameters, achieving flexible and interpretable domain adaptation at inference time. Evaluated on regression tasks including compressive sensing, gain calibration, and phase retrieval, the proposed approaches significantly outperform joint-training baselines and match or even surpass the performance of domain-specific models, effectively balancing generality and accuracy.

Abstract

Machine learning models often struggle to generalize across domains with varying data distributions, such as differing noise levels, leading to degraded performance. Traditional strategies like personalized training, which trains separate models per domain, and joint training, which uses a single model for all domains, have significant limitations in flexibility and effectiveness. To address this, we propose two novel domain adaptation methods for regression tasks based on interpretable unrolled networks--deep architectures inspired by iterative optimization algorithms. These models leverage the functional dependence of select tunable parameters on domain variables, enabling controlled adaptation during inference. Our methods include Parametric Tunable-Domain Adaptation (P-TDA), which uses known domain parameters for dynamic tuning, and Data-Driven Tunable-Domain Adaptation (DD-TDA), which infers domain adaptation directly from input data. We validate our approach on compressed sensing problems involving noise-adaptive sparse signal recovery, domain-adaptive gain calibration, and domain-adaptive phase retrieval, demonstrating improved or comparable performance to domain-specific models while surpassing joint training baselines. This work highlights the potential of unrolled networks for effective, interpretable domain adaptation in regression settings.

Problem

• domain adaptation
• generalization
• data distribution shift
• regression tasks
• noise levels

Innovation

• Unrolled Networks
• Domain Adaptation
• Tunable Parameters
• Regression
• Interpretable AI