Research map / Adversarial robustness and security / Certified and empirical robustness
Verifying Properties of Binarized Deep Neural Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Replace the ILP encoding with the proposed SAT encoding | 4-block BNN robustness verification · MNIST, MNIST-rot, and MNIST-back-image at ε=1, 3, and 5 | instances solved: 180 vs 130; 179 vs 125; 191 vs 143 | Helped |
| Add counterexample-guided search to the SAT encoding | 4-block BNN robustness verification · MNIST, MNIST-rot, and MNIST-back-image | instances solved: MNIST ε=1: 180 vs 171; MNIST-rot ε=1: 179 vs 197; MNIST-back-image ε=1: 191 vs 191 | Mixed |
| Add counterexample-guided search to the SAT encoding | 4-block BNN robustness verification · MNIST, MNIST-rot, and MNIST-back-image | average solving time (seconds): 77.3 -> 34.1; 57.4 -> 13.5; 18.3 -> 12.8 | Helped |
| Increase the allowed perturbation bound ε | 4-block BNN on MNIST-back-image · MNIST-back-image | certifiably ε-robust images: 138 -> 20 -> 3 | Hurt |
| Add input batch normalization and binarization layers | BNN input processing · MNIST, MNIST-rot, and MNIST-back-image | classification accuracy: less than one percent | Hurt |
Closest papers by meaning
- BOLD: Boolean Logic Deep Learning (NeurIPS 2024 · similarity 0.88)
- Verifying Properties of Binary Neural Networks Using Sparse Polynomial Optimization (ICLR 2025 · similarity 0.88)
- VNN: Verification-Friendly Neural Networks with Hard Robustness Guarantees (ICML 2024 · similarity 0.85)
- Certified but Private: Scalable Zero-Knowledge Proofs for the Formal Verification of Neural Networks (NeurIPS 2026 · similarity 0.85)
- Interpreting Robustness Proofs of Deep Neural Networks (ICLR 2024 · similarity 0.85)
- Towards Efficient Verification of Quantized Neural Networks (AAAI 2024 · similarity 0.84)
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.