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 [1].
Beyond the Golden Ratio for Variational Inequality Algorithms
Authors: Ahmet Alacaoglu, Axel Böhm, Yura Malitsky
JMLR 2023 | Venue PDF | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | with superior empirical performance. Keywords: min-max, variational inequality, adaptive step size, nonmonotone; 5.3 Experiments |
| Researcher Affiliation | Academia | Ahmet Alacaoglu1 EMAIL Wisconsin Institute for Discovery University of Wisconsin Madison Madison, WI, USA Axel Böhm1 EMAIL Faculty of Mathematics University of Vienna Vienna, Austria Yura Malitsky1 EMAIL Faculty of Mathematics University of Vienna Vienna, Austria |
| Pseudocode | Yes | Algorithm 1 a GRAAL; Algorithm 2 GRAAL |
| Open Source Code | No | The paper describes algorithms (a GRAAL and Curvature EG+) and their parameters for experiments but does not provide any statement or link about the availability of the authors' source code for these methodologies. |
| Open Datasets | Yes | For the left plot in Fig. 2 we use the Lagrangian formulation of a linearly constrained quadratic program ... We use the parametrization proposed in (Ouyang and Xu, 2021) and further studied in (Yoon and Ryu, 2021)... The middle and right plot in Fig. 2 are special instances of bilinear matrix games... given in (Nemirovski et al., 2009). Figure 3: A special parametrization of the Polar Game example from (Pethick et al., 2022)... Figure 4: The Forsaken example of (Hsieh et al., 2021, Example 5.2)... |
| Dataset Splits | No | The paper uses mathematical formulations of problems like linearly constrained quadratic programs and matrix games for experiments, which do not involve explicit training/test/validation dataset splits. No specific dataset split information is provided. |
| Hardware Specification | No | The paper describes experimental results and algorithms but does not specify the hardware (e.g., CPU, GPU models, memory) used for running the experiments. |
| Software Dependencies | No | The paper describes the algorithms used (a GRAAL, Curvature EG+) and their parameters, but it does not specify any software dependencies (e.g., programming languages, libraries, or solvers) with version numbers. |
| Experiment Setup | Yes | For (a GRAAL) φ is usually given in the legend, except for Fig. 1 where we used the default φ = 1.5. If γ is not given in the legend we use the theoretical upper bound 1/φ + 1/φ2. For Curvature EG+ we use a backtracking linesearch initialized with ν JF(zk) 1, where we use ν = 0.99 and JF denotes the Jacobian of F. ... the step size is decreased by τ = 0.9. |