SEEV: Synthesis with Efficient Exact Verification for ReLU Neural Barrier Functions

Authors: Hongchao Zhang, Zhizhen Qin, Sicun Gao, Andrew Clark

NeurIPS 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our simulations show that SEEV significantly improves verification efficiency while maintaining the CBF quality across various benchmark systems and neural network structures. Our simulation results demonstrate significant improvements in verification efficiency and reliability across a range of benchmark systems. In this section, we evaluate the proposed SEEV in regularizer efficacy and verification efficiency. We also demonstrated the improved performance with counter-example guidance in the synthesis framework, whose results are detailed in B.2. We experiment on four systems, namely Darboux, obstacle avoidance, hi-ord8, and spacecraft rendezvous.
Researcher Affiliation Academia Hongchao Zhang Electrical & Systems Engineering Washington University in St. Louis hongchao@wustl.edu Zhizhen Qin Computer Science & Engineering University of California, San Diego zhizhenqin@ucsd.edu Sicun Gao Computer Science & Engineering University of California, San Diego scungao@ucsd.edu Andrew Clark Electrical & Systems Engineering Washington University in St. Louis andrewclark@wustl.edu
Pseudocode Yes Algorithm 1 Efficient Exact Verification; Algorithm 2 Binary Search for S0; Algorithm 3 Enumerate Activated Sets; Algorithm 4 Enumerate Hinges
Open Source Code Yes Our code is available at https://github.com/Hongchao Zhang-HZ/SEEV.
Open Datasets No The training dataset T is initialized by uniform sampling over X.
Dataset Splits No The paper initializes its training dataset by uniform sampling and updates it with counterexamples, but does not define fixed train/validation/test dataset splits with percentages or counts.
Hardware Specification Yes The experiments run on a workstation with an Intel i7-11700KF CPU, and an NVIDIA Ge Force RTX 3080 GPU.
Software Dependencies No The paper mentions tools like 'd Real' and 'Z3' but does not specify version numbers for general software dependencies or libraries.
Experiment Setup Yes We include experiment settings in Appendix B.1 and hyperparameters settings in Table 4 in Appendix B.3.