Research map / Learning theory and algorithms / Combinatorial and search optimization
Provably Bounding Neural Network Preimages
Suhas Kotha, Christopher Brix, J Zico Kolter, Krishnamurthy Dj Dvijotham, Huan Zhang
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 |
|---|---|---|---|
| INVPROP (output-constraint-aware bound propagation) instead of ReBreach-LP | backward reachability · Double Integrator | approximation ratio: 4021.47 -> 1.46 | Improved |
| INVPROP instead of ReBreach-LP | backward reachability · Double Integrator | time (sec): 42.86 -> 17.89 | Improved |
| INVPROP instead of ReBreach-LP with no partitions | backward reachability · 6D Quadrotor | approx volume: 64 -> 0.0000249 | Improved |
| INVPROP instead of ReBreach-LP with 15625 output partitions | backward reachability · 6D Quadrotor | approx volume: 0.064 -> 0.0000249 | Improved |
| INVPROP instead of ReBreach-LP with 15625 output partitions | backward reachability · 6D Quadrotor | time (sec): 662.2 -> 213.8 | Improved |
| INVPROP instead of ReBreach-LP with no partitions | backward reachability · 6D Quadrotor | time (sec): 11.8 -> 213.8 | Hurt |
| add INVPROP to alpha,beta-CROWN | alpha,beta-CROWN · MNIST FFN4 (3x100) | verified instances: 45 -> 45 | No clear effect |
| add INVPROP to alpha,beta-CROWN | alpha,beta-CROWN · MNIST FFN5 (6x100) | verified instances: 38 -> 39 | Improved |
| add INVPROP to alpha,beta-CROWN | alpha,beta-CROWN · MNIST FFN6 (6x200) | avg time [s]: 5.4 -> 1.0 | Improved |
| add INVPROP to alpha,beta-CROWN | alpha,beta-CROWN · MNIST FFN7 (9x200) | verified instances: 37 -> 40 | Improved |
| add INVPROP to alpha,beta-CROWN | alpha,beta-CROWN · MNIST FFN5 (6x100) | avg time [s]: 6.7 -> 8.7 | Hurt |
| add INVPROP on last two intermediate layers | alpha,beta-CROWN · VNN-COMP 2023 YOLO | verified instances: +6 verified | Improved |
Closest papers by meaning
- Clip-and-Verify: Linear Constraint-Driven Domain Clipping for Accelerating Neural Network Verification (NeurIPS 2025 · similarity 0.88)
- Certified but Private: Scalable Zero-Knowledge Proofs for the Formal Verification of Neural Networks (NeurIPS 2026 · similarity 0.88)
- VNN: Verification-Friendly Neural Networks with Hard Robustness Guarantees (ICML 2024 · similarity 0.88)
- Dynamic Algorithm Termination for Branch-and-Bound-based Neural Network Verification (AAAI 2025 · similarity 0.88)
- Understanding Certified Training with Interval Bound Propagation (ICLR 2024 · similarity 0.88)
- SDP-CROWN: Efficient Bound Propagation for Neural Network Verification with Tightness of Semidefinite Programming (ICML 2025 · similarity 0.88)
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.
