DP-TTA: Test-time Adaptation for Transient Electromagnetic Signal Denoising via Dictionary-driven Prior Regularization

2025-10-15

AI Summary

Transient electromagnetic (TEM) signals in real-world scenarios are corrupted by geographically dependent noise; existing deep learning denoising models exhibit poor cross-regional generalization due to reliance on synthetic or single-source training data. To address this, we propose a dictionary-driven prior-regularized test-time adaptation method: for the first time, we encode physics-based priors—such as exponential decay and smoothness of TEM signals—into a learnable dictionary, and dynamically optimize network parameters during inference via a self-supervised loss combining dictionary consistency and first-order variation constraints. The method requires no target-domain labels, leveraging only intrinsic physical properties of the signal for environment-aware adaptation. Experiments demonstrate that our approach significantly outperforms state-of-the-art denoising models and existing test-time adaptation methods across diverse real-world noisy scenarios, achieving substantial improvements in signal-to-noise ratio and structural similarity index, thereby enhancing cross-scenario robustness and signal reconstruction fidelity.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Search and Optimization: Learning to Search
• Natural Language Processing: Learning & Optimization for NLP

Application Category

• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Graph Algorithms and Modeling for the Web: Graph neural networks and deep learning approaches for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Transient Electromagnetic (TEM) method is widely used in various geophysical applications, providing valuable insights into subsurface properties. However, time-domain TEM signals are often submerged in various types of noise. While recent deep learning-based denoising models have shown strong performance, these models are mostly trained on simulated or single real-world scenario data, overlooking the significant differences in noise characteristics from different geographical regions. Intuitively, models trained in one environment often struggle to perform well in new settings due to differences in geological conditions, equipment, and external interference, leading to reduced denoising performance. To this end, we propose the Dictionary-driven Prior Regularization Test-time Adaptation (DP-TTA). Our key insight is that TEM signals possess intrinsic physical characteristics, such as exponential decay and smoothness, which remain consistent across different regions regardless of external conditions. These intrinsic characteristics serve as ideal prior knowledge for guiding the TTA strategy, which helps the pre-trained model dynamically adjust parameters by utilizing self-supervised losses, improving denoising performance in new scenarios. To implement this, we customized a network, named DTEMDNet. Specifically, we first use dictionary learning to encode these intrinsic characteristics as a dictionary-driven prior, which is integrated into the model during training. At the testing stage, this prior guides the model to adapt dynamically to new environments by minimizing self-supervised losses derived from the dictionary-driven consistency and the signal one-order variation. Extensive experimental results demonstrate that the proposed method achieves much better performance than existing TEM denoising methods and TTA methods.

Problem

• Addressing noise variation across geographical regions in TEM denoising
• Adapting pre-trained models to new environments via physical priors
• Improving denoising performance using dictionary-driven consistency and signal variation

Innovation

• Dictionary-driven prior regularization for test-time adaptation
• Self-supervised losses from dictionary consistency and signal variation
• Dynamic parameter adjustment of pre-trained model for new scenarios