Enhancing Robustness of Implicit Neural Representations Against Weight Perturbations

2025-08-18

AI Summary

Implicit Neural Representations (INRs) suffer significant degradation in reconstruction quality under weight perturbations, revealing poor robustness. This work presents the first systematic investigation into the sensitivity mechanisms of INRs to weight perturbations and proposes a gradient-regularized robust training framework. Specifically, it minimizes the discrepancy between reconstruction losses before and after perturbation while explicitly constraining the magnitude of the loss gradient with respect to network weights—thereby enhancing model stability without increasing inference overhead or sacrificing architectural compatibility with mainstream INR designs. Evaluated across diverse modalities—including image, video, and 3D scene reconstruction—the method achieves up to a 7.5 dB PSNR improvement under noise corruption, substantially outperforming standard INRs. This establishes a novel paradigm for robust implicit modeling.

Technology Category

• Computer Vision: Adversarial Attacks & Robustness
• Machine Learning: Adversarial Learning & Robustness
• Natural Language Processing: Safety and Robustness

Application Category

• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Responsible Web: Machine-in-the-loop, human agency and autonomy
• Graph Algorithms and Modeling for the Web: Graph embeddings and representation learning for Web-related graphs

Abstract

Implicit Neural Representations (INRs) encode discrete signals in a continuous manner using neural networks, demonstrating significant value across various multimedia applications. However, the vulnerability of INRs presents a critical challenge for their real-world deployments, as the network weights might be subjected to unavoidable perturbations. In this work, we investigate the robustness of INRs for the first time and find that even minor perturbations can lead to substantial performance degradation in the quality of signal reconstruction. To mitigate this issue, we formulate the robustness problem in INRs by minimizing the difference between loss with and without weight perturbations. Furthermore, we derive a novel robust loss function to regulate the gradient of the reconstruction loss with respect to weights, thereby enhancing the robustness. Extensive experiments on reconstruction tasks across multiple modalities demonstrate that our method achieves up to a 7.5~dB improvement in peak signal-to-noise ratio (PSNR) values compared to original INRs under noisy conditions.

Problem

• Enhancing robustness of Implicit Neural Representations against weight perturbations
• Addressing vulnerability of INRs to minor weight changes causing performance degradation
• Mitigating substantial signal reconstruction quality loss under noisy conditions

Innovation

• Minimizing loss difference with perturbations
• Novel robust loss function regulating gradients
• Enhancing robustness against weight perturbations