On the Paradox of Certified Training
Nikola Jovanovi\'c, Mislav Balunovi\'c, Maximilian Baader, Martin, Vechev

TL;DR
This paper investigates why looser convex relaxations sometimes yield higher certified robustness in training models, revealing that properties like continuity and sensitivity significantly influence training dynamics and robustness outcomes.
Contribution
It identifies key properties beyond tightness, such as continuity and sensitivity, that affect certified training, and systematically explores their impact and tradeoffs.
Findings
Looser relaxations can outperform tighter ones in certified robustness.
Continuity and sensitivity of relaxations influence training effectiveness.
Improving relaxation properties often involves complex tradeoffs.
Abstract
Certified defenses based on convex relaxations are an established technique for training provably robust models. The key component is the choice of relaxation, varying from simple intervals to tight polyhedra. Counterintuitively, loose interval-based training often leads to higher certified robustness than what can be achieved with tighter relaxations, which is a well-known but poorly understood paradox. While recent works introduced various improvements aiming to circumvent this issue in practice, the fundamental problem of training models with high certified robustness remains unsolved. In this work, we investigate the underlying reasons behind the paradox and identify two key properties of relaxations, beyond tightness, that impact certified training dynamics: continuity and sensitivity. Our extensive experimental evaluation with a number of popular convex relaxations provides strong…
Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
TopicsAdversarial Robustness in Machine Learning · Formal Methods in Verification · Explainable Artificial Intelligence (XAI)
