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..
Semi-infinite Nonconvex Constrained Min-Max Optimization
Authors: Cody Melcher, Zeinab Alizadeh, Lindsey Hiett, Afrooz Jalilzadeh, Erfan Yazdandoost Hamedani
NeurIPS 2025 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Numerical experiments on robust multitask learning with task priority further illustrate the practical effectiveness of the algorithm. In this section, we evaluate the performance of the proposed i DB-PD algorithm on the Robust Multi Task 2 problem with a task priority, as introduced in Section 1.2. All experiments were implemented in PyTorch and executed on Google Colab, using a virtual machine equipped with an NVIDIA A100-SXM4 GPU (40 GB), an Intel Xeon CPU @ 2.20 GHz, 87 GB of RAM, and running Ubuntu 22.04.4 LTS with Python 3.12. |
| Researcher Affiliation | Academia | Cody Melcher School of Mathematical Sciences The University of Arizona Tucson, AZ 85721 EMAIL Zeinab Alizadeh Systems and Industrial Engineering The University of Arizona Tucson, AZ 85721 EMAIL Lindsey Heitt Systems and Industrial Engineering The University of Arizona Tucson, AZ 85721 EMAIL Afrooz Jalilzadeh Systems and Industrial Engineering The University of Arizona Tucson, AZ 85721 EMAIL Erfan Yazdandoost Hamedani Systems and Industrial Engineering The University of Arizona Tucson, AZ 85721 EMAIL |
| Pseudocode | Yes | Algorithm 1 Inexact Dynamic Barrier Primal-Dual (i DB-PD) Method for Semi-Infinite Min-Max |
| Open Source Code | Yes | 5. Open access to data and code Question: Does the paper provide open access to the data and code, with sufficient instructions to faithfully reproduce the main experimental results, as described in supplemental material? Answer: [Yes] Justification: The code and data are attached in the supplementary material. |
| Open Datasets | Yes | We used Multi-MNIST and Multi-Fashion-MNIST from Lin et al. (2019) [44]1, which were constructed from the original MNIST dataset. ... Yeast, Coronary Heart Disease, and 20News Group datasets were accessed via the Multi-Label Classification Dataset Repository hosted by Universidad de Córdoba2. 1Datasets by Lin et al. (2019): https://github.com/Xi-L/Pareto MTL/ 2https://www.uco.es/kdis/mllresources/ |
| Dataset Splits | No | The paper mentions that for the Multi-MNIST and Multi-Fashion-MNIST datasets, each data point is constructed by randomly sampling two different images and combining them. For all datasets, it states, "we evenly partition the labels into two disjoint subsets". However, it does not provide specific training, validation, or test dataset splits (e.g., percentages or counts of samples for each split) for experimental reproduction. |
| Hardware Specification | Yes | All experiments were implemented in PyTorch and executed on Google Colab, using a virtual machine equipped with an NVIDIA A100-SXM4 GPU (40 GB), an Intel Xeon CPU @ 2.20 GHz, 87 GB of RAM, and running Ubuntu 22.04.4 LTS with Python 3.12. |
| Software Dependencies | Yes | All experiments were implemented in PyTorch and executed on Google Colab, using a virtual machine equipped with an NVIDIA A100-SXM4 GPU (40 GB), an Intel Xeon CPU @ 2.20 GHz, 87 GB of RAM, and running Ubuntu 22.04.4 LTS with Python 3.12. |
| Experiment Setup | Yes | Experiment Details: In all experiments, we select the regularization parameter λ = 10^-3 and the maximization variables y, w are updated by running Nk = 2 log(k + 2) and Mk = 10 log(k + 2) steps of the projected gradient ascent method. The stepsize γ is tuned by selecting the best performance among {10^-4, 2.5 10^-4, 5 10^-4, 10^-3, 5 10^-3, 10^-2} and the parameter is set αk = α/(k + 2)^1.001 for α {0.1, 0.2, 0.5, 1}. |