Predictive Spectral Calibration for Source-Free Test-Time Regression

2026-03-10

AI Summary

This work addresses the performance degradation of image regression models under distribution shifts at test time in source-free scenarios. To tackle this challenge, the authors propose a novel test-time adaptation framework that operates without access to source data. The method extends subspace alignment to the block spectral matching level by jointly aligning target features in both the source prediction-supporting subspace and its orthogonal complement. Furthermore, it introduces a model-agnostic residual spectral calibration mechanism to effectively adapt to continuous regression targets. Evaluated on multiple image regression benchmarks, the proposed approach significantly outperforms strong existing baselines, demonstrating particularly robust performance under severe distribution shifts.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Adversarial Attacks & Robustness
• Intelligent Robots: Multimodal Perception & Sensor Fusion

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications
• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs

Abstract

Test-time adaptation (TTA) for image regression has received far less attention than its classification counterpart. Methods designed for classification often depend on classification-specific objectives and decision boundaries, making them difficult to transfer directly to continuous regression targets. Recent progress revisits regression TTA through subspace alignment, showing that simple source-guided alignment can be both practical and effective. Building on this line of work, we propose Predictive Spectral Calibration (PSC), a source-free framework that extends subspace alignment to block spectral matching. Instead of relying on a fixed support subspace alone, PSC jointly aligns target features within the source predictive support and calibrates residual spectral slack in the orthogonal complement. PSC remains simple to implement, model-agnostic, and compatible with off-the-shelf pretrained regressors. Experiments on multiple image regression benchmarks show consistent improvements over strong baselines, with particularly clear gains under severe distribution shifts.

Problem

• test-time adaptation
• image regression
• source-free
• distribution shift
• subspace alignment

Innovation

• Predictive Spectral Calibration
• Test-Time Adaptation
• Source-Free Regression
• Subspace Alignment
• Spectral Matching