Understanding GNNs for Boolean Satisfiability through Approximation Algorithms
This work addresses the problem of understanding GNN decision-making for researchers in AI and theoretical computer science, though it is incremental as it builds on existing approximation algorithms.
The paper tackled the interpretability of Graph Neural Networks (GNNs) for Boolean Satisfiability by connecting them to approximation algorithms like Belief Propagation and Semidefinite Programming Relaxations, resulting in a curriculum training procedure that reduced training time by more than an order of magnitude and increased the percentage of solved problems.
The paper deals with the interpretability of Graph Neural Networks in the context of Boolean Satisfiability. The goal is to demystify the internal workings of these models and provide insightful perspectives into their decision-making processes. This is done by uncovering connections to two approximation algorithms studied in the domain of Boolean Satisfiability: Belief Propagation and Semidefinite Programming Relaxations. Revealing these connections has empowered us to introduce a suite of impactful enhancements. The first significant enhancement is a curriculum training procedure, which incrementally increases the problem complexity in the training set, together with increasing the number of message passing iterations of the Graph Neural Network. We show that the curriculum, together with several other optimizations, reduces the training time by more than an order of magnitude compared to the baseline without the curriculum. Furthermore, we apply decimation and sampling of initial embeddings, which significantly increase the percentage of solved problems.