Adaptive Branch-and-Bound Tree Exploration for Neural Network Verification
By: Kota Fukuda , Guanqin Zhang , Zhenya Zhang and more
Potential Business Impact:
Finds mistakes in smart computer brains faster.
Formal verification is a rigorous approach that can provably ensure the quality of neural networks, and to date, Branch and Bound (BaB) is the state-of-the-art that performs verification by splitting the problem as needed and applying off-the-shelf verifiers to sub-problems for improved performance. However, existing BaB may not be efficient, due to its naive way of exploring the space of sub-problems that ignores the \emph{importance} of different sub-problems. To bridge this gap, we first introduce a notion of ``importance'' that reflects how likely a counterexample can be found with a sub-problem, and then we devise a novel verification approach, called ABONN, that explores the sub-problem space of BaB adaptively, in a Monte-Carlo tree search (MCTS) style. The exploration is guided by the ``importance'' of different sub-problems, so it favors the sub-problems that are more likely to find counterexamples. As soon as it finds a counterexample, it can immediately terminate; even though it cannot find, after visiting all the sub-problems, it can still manage to verify the problem. We evaluate ABONN with 552 verification problems from commonly-used datasets and neural network models, and compare it with the state-of-the-art verifiers as baseline approaches. Experimental evaluation shows that ABONN demonstrates speedups of up to $15.2\times$ on MNIST and $24.7\times$ on CIFAR-10. We further study the influences of hyperparameters to the performance of ABONN, and the effectiveness of our adaptive tree exploration.
Similar Papers
BaB-prob: Branch and Bound with Preactivation Splitting for Probabilistic Verification of Neural Networks
Machine Learning (CS)
Checks if AI makes the right choices.
Clip-and-Verify: Linear Constraint-Driven Domain Clipping for Accelerating Neural Network Verification
Machine Learning (CS)
Makes AI smarter and safer by checking its work.
Non-linear Multi-objective Optimization with Probabilistic Branch and Bound
Optimization and Control
Finds best choices when results are uncertain.