BDD4BNN: A BDD-based Quantitative Analysis Framework for Binarized Neural Networks
This provides a quantitative verification framework for BNNs, addressing safety-critical applications, but it is incremental as it builds on existing BDD methods for neural networks.
The paper tackles the verification of Binarized Neural Networks (BNNs) by encoding them into Binary Decision Diagrams (BDDs) to enable precise analysis, and demonstrates effectiveness through experiments with a prototype tool.
Verifying and explaining the behavior of neural networks is becoming increasingly important, especially when they are deployed in safety-critical applications. In this paper, we study verification problems for Binarized Neural Networks (BNNs), the 1-bit quantization of general real-numbered neural networks. Our approach is to encode BNNs into Binary Decision Diagrams (BDDs), which is done by exploiting the internal structure of the BNNs. In particular, we translate the input-output relation of blocks in BNNs to cardinality constraints which are then encoded by BDDs. Based on the encoding, we develop a quantitative verification framework for BNNs where precise and comprehensive analysis of BNNs can be performed. We demonstrate the application of our framework by providing quantitative robustness analysis and interpretability for BNNs. We implement a prototype tool BDD4BNN and carry out extensive experiments which confirm the effectiveness and efficiency of our approach.