On Representing Convex Quadratically Constrained Quadratic Programs via Graph Neural Networks
Authors: Chenyang Wu, Qian Chen, Akang Wang, Tian Ding, Ruoyu Sun, Wenguo Yang, Qingjiang Shi
TMLR 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We conduct initial numerical tests of the tripartite MP-GNNs on small QCQP instances. The results showed that MP-GNNs can be trained to approximate the key properties well. ... In this section, we present empirical experiments to validate the proposed theoretical results. |
| Researcher Affiliation | Academia | 1University of Chinese Academy of Sciences, China 2Shenzhen Research Institute of Big Data, China 3School of Science and Engineering, The Chinese University of Hong Kong, Shenzhen, China 4School of Data Science, The Chinese University of Hong Kong, Shenzhen, China 5School of Computer Science and Technology, Tongji University, Shanghai, China |
| Pseudocode | No | The paper describes the GNN architecture and message-passing layers using mathematical notation and a diagram (Figure 2), but does not contain explicitly labeled pseudocode or algorithm blocks with structured steps. |
| Open Source Code | Yes | The corresponding source code is available at https://github.com/Net Sys Opt/L2QP. |
| Open Datasets | Yes | We incorporated a dataset derived from real-world instances in QPLib. Due to computational constraints, we used all instances in QPLIB that are convex and have no more than 5,000 nonzeros as training sets, and augmented these instances by perturbations. ... Qplib: a library of quadratic programming instances. Mathematical Programming Computation, 11:237 265, 2019. |
| Dataset Splits | Yes | For each dataset, we generate 1,000 instances for training and 300 instances for validation. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., GPU/CPU models, processor types) used for running its experiments. It only generally refers to 'runtime comparison' without specifying the computational environment. |
| Software Dependencies | Yes | All instances are solved using the IPOPT solver (Wächter & Biegler, 2006). ... Gurobi (Gurobi Optimization, LLC, 2025), MOSEK (Ap S, 2025), SCS (O Donoghue et al., 2016), and ECOS (Domahidi et al., 2013). ... MOSEK Ap S. The MOSEK Python Fusion API manual. Version 11.0., 2025. |
| Experiment Setup | Yes | The hyper-parameter T is set to 2. For training, we utilized the Adam optimizer with a learning rate of 0.0001 and a batch size of 16. |