Verifying Quantized Graph Neural Networks is PSPACE-complete

Authors: Marco Sälzer, Francois Schwarzentruber, Nicolas Troquard

IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We introduce the linear-constrained validity (LVP) problem for verifying GNNs properties, and provide an efficient translation from LVP instances into a logical language. We show that LVP is in PSPACE, for any reasonable activation functions. We provide a proof system. We also prove PSPACE-hardness, indicating that while reasoning about quantized GNNs is feasible, it remains generally computationally challenging.
Researcher Affiliation Academia 1 Technical University of Kaiserslautern, Kaiserslautern, Germany 2ENS de Lyon, CNRS, Universit e Claude Bernard Lyon 1, Inria, LIP, UMR 5668, 69342, Lyon, France 3Gran Sasso Science Institute (GSSI), Viale F. Crispi, 7 67100 L Aquila, Italy
Pseudocode Yes Figure 4: Tableau rules for the satisfiability problem in Lquant GNN when δ is written in unary.
Open Source Code Yes A prototype implementation1 of the proof system is presented in [S alzer et al., 2025]. 1https://github.com/francoisschwarzentruber/ijcai2025-verifquantgnn
Open Datasets No The paper focuses on theoretical verification of GNNs and introduces a logical framework. It does not perform experiments on specific datasets and therefore does not provide access information for any.
Dataset Splits No The paper is theoretical and does not involve experimental evaluation on datasets. Therefore, it does not describe dataset splits.
Hardware Specification No The paper is purely theoretical, focusing on logical frameworks and complexity proofs for GNN verification. It does not describe any experiments that would require specific hardware specifications.
Software Dependencies No The paper mentions a prototype implementation of the proof system in Python via a GitHub link, but it does not specify version numbers for Python or any specific libraries/solvers used in that implementation. It mentions PyTorch Geometric as a third-party library for general GNNs but not as a specific dependency for its own methodology with version.
Experiment Setup No The paper presents theoretical work on verifying quantized GNNs and does not include experimental results or a detailed experimental setup with hyperparameters or training configurations.