Research map / Learning theory and algorithms / Prediction and decision losses
A Robust Optimisation Perspective on Counterexample-Guided Repair of Neural Networks
David Boetius, Stefan Leue, Tobias Sutter
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Solve the linear-regression repair problem directly as a quadratic program | RMI second-stage linear regression · RMI second-stage linear regression, ε = 100 | success rate: 58% -> 72 % | Improved |
| Solve the linear-regression repair problem directly as a quadratic program | RMI second-stage linear regression · RMI second-stage linear regression, ε = 150 | success rate: 94% -> 97 % | Improved |
| Use an early-exit verifier instead of an optimal verifier | Counterexample-guided repair · MNIST and ACAS Xu | runtime | Improved |
| Use an optimal verifier instead of an early-exit verifier | Counterexample-guided repair · MNIST and ACAS Xu | repair steps | Improved |
| Use an early-exit verifier instead of an optimal verifier | Counterexample-guided repair · CollisionDetection and RMI first-stage networks | runtime | No clear effect |
| Use BIM falsification to search for counterexamples before invoking the verifier | MNIST repair · MNIST | runtime | Improved |
Closest papers by meaning
- Sound and Complete Neural Network Repair with Minimality and Locality Guarantees (ICLR 2022 · similarity 0.89)
- REGLO: Provable Neural Network Repair for Global Robustness Properties (AAAI 2024 · similarity 0.86)
- Overcoming the Convex Barrier for Simplex Inputs (NeurIPS 2021 · similarity 0.85)
- Formalising the Robustness of Counterfactual Explanations for Neural Networks (AAAI 2023 · similarity 0.85)
- Repairing Neural Networks by Leaving the Right Past Behind (NeurIPS 2022 · similarity 0.85)
- Conserve-Update-Revise to Cure Generalization and Robustness Trade-off in Adversarial Training (ICLR 2024 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
