Exploring the Vulnerability of Deep Neural Networks: A Study of Parameter Corruption
arXiv:2006.05620
Abstract
We argue that the vulnerability of model parameters is of crucial value to the study of model robustness and generalization but little research has been devoted to understanding this matter. In this work, we propose an indicator to measure the robustness of neural network parameters by exploiting their vulnerability via parameter corruption. The proposed indicator describes the maximum loss variation in the non-trivial worst-case scenario under parameter corruption. For practical purposes, we give a gradient-based estimation, which is far more effective than random corruption trials that can hardly induce the worst accuracy degradation. Equipped with theoretical support and empirical validation, we are able to systematically investigate the robustness of different model parameters and reveal vulnerability of deep neural networks that has been rarely paid attention to before. Moreover, we can enhance the models accordingly with the proposed adversarial corruption-resistant training, which not only improves the parameter robustness but also translates into accuracy elevation.
arXiv commentAccepted by AAAI 2021
References in corpus (9)
- How transferable are features in deep neural networks?
- Targeted Backdoor Attacks on Deep Learning Systems Using Data Poisoning
- Pose Guided Person Image Generation
- Regularizing and Optimizing LSTM Language Models
- Understanding Adversarial Training: Increasing Local Stability of Neural Nets through Robust Optimization
- Real-valued (Medical) Time Series Generation with Recurrent Conditional GANs
- fairseq: A Fast, Extensible Toolkit for Sequence Modeling
- An Analysis of Neural Language Modeling at Multiple Scales
- LCA: Loss Change Allocation for Neural Network Training
Cited by in corpus (5)
- Sharpness-Aware Minimization for Efficiently Improving Generalization
- ASAM: Adaptive Sharpness-Aware Minimization for Scale-Invariant Learning of Deep Neural Networks
- Raise a Child in Large Language Model: Towards Effective and Generalizable Fine-tuning
- Minimum sharpness: Scale-invariant parameter-robustness of neural networks
- Relative Flatness and Generalization