Notice: The reproducibility variables underlying each score are classified using an automated LLM-based pipeline, validated against a manually labeled dataset. LLM-based classification introduces uncertainty and potential bias; scores should be interpreted as estimates. Full accuracy metrics and methodology are described in Coakley et alK. L. Coakley, T. Snelleman, H. Hoos, and O. E. Gundersen, "The embrace of open science: An analysis of a decade of AI research and 56 800 conference papers," Under Review, 2026..
Smart Initial Basis Selection for Linear Programs
Authors: Zhenan Fan, Xinglu Wang, Oleksandr Yakovenko, Abdullah Ali Sivas, Owen Ren, Yong Zhang, Zirui Zhou
ICML 2023 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We validate the effectiveness of our proposed strategy by extensively testing it with state-of-the-art simplex solvers, including the open-source solver Hi GHS and the commercial solver Opt Verse. Through these rigorous experiments, we demonstrate that our strategy achieves substantial speedup and consistently outperforms existing rule-based methods. |
| Researcher Affiliation | Collaboration | 1Huawei Technologies Canada, Burnaby, Canada 2Simon Fraser University, Burnaby, Canada. |
| Pseudocode | Yes | Algorithm 1 Smart Initial Basis Selection Algorithm for Linear Programs |
| Open Source Code | Yes | Our code is publicly available at Huawei AI Gallery 1. 1https://developer.huaweicloud.com/develop/aigallery/notebook/detail?id=ce45dd10-44ce-43bb-89c8-1f3277f1132d |
| Open Datasets | Yes | The publicly available datasets include the Maritime Inventory Routing Problems (MIRP) (Papageorgiou et al., 2014), the One-norm Support Vector Machine Instances (LIBSVM) (Zhu et al., 2003; Applegate et al., 2021), and the 2-stage Stochastic Problems (STOCH) (Castro & de la Lama-Zubir an, 2020). |
| Dataset Splits | Yes | To ensure a fair evaluation, each dataset is split into training and test sets in a 7:3 ratio. |
| Hardware Specification | Yes | The GNN model is trained on NVIDIA V100. The evaluation is conducted on a system comprised of 8 cores CPU (Intel Xeon E5-2690 v4) and 64 G of memory, utilizing Ubuntu 18.04 Docker containers for solver execution. |
| Software Dependencies | Yes | We implement our approach with Python 3.7, Py Torch 1.8 and Py G framework (Fey & Lenssen, 2019). |
| Experiment Setup | Yes | By default, 3 layers lightweight GNN is used for SC-1 and SC-2 datasets, and 5 layers GNN is used for other datasets. The size of the hidden layers is 128 and the dropout ratio is 0.1. ... We do not tune hyper-parameter for training GNN and just adopt the commonly used one, like the Adam optimizer with initial learning rate 1e-3 and weight decay 1e-4. The learning rate decays by 0.1 every 200 epochs and the total training epoch is 800. |