- Rating
- 1
- Confidence
- 3
- Informativeness
- 1
- Presentation
- 2
- Soundness
- 2
Read review
Paper summary
This paper studies why adversarial training improves clean forecasting performance for nonlinear models but hurts compact linear models. It attributes this difference to the trade-off between prediction-variance reduction and parameter displacement. The author proposes DGA, which anchors adversarially trained parameters to a clean ERM checkpoint using an penalty.
Critical strengths
- This paper reveals a capacity-dependent reversal in the effect of adversarial training on clean performance.
- The experiments cover multiple architectures, datasets, forecasting horizons, and random seeds.
Critical weaknesses
- The novelty is limited because DGA is essentially adversarial training with an penalty toward a clean ERM checkpoint, closely related to existing parameter-anchoring methods.
- The comparison may be unfair because DGA searches five anchoring coefficients over five seeds, while ERM and other regularization baselines are not given an equivalent tuning budget.
- Selecting DGA from twenty-five training runs may introduce a substantial validation-selection advantage.
- The clean-only anchored baseline performs slightly better than DGA in the diagnostic linear settings, making it unclear whether adversarial perturbations are necessary.
- The method has high training cost, some improvements are small.
Questions for authors
Besides the weakness section, another concern is that the paper’s figures and text appear to have been excessively processed by LLMs.