Game Theoretic Optimization via Gradient-based Nikaido-Isoda Function
Authors: Arvind Raghunathan, Anoop Cherian, Devesh Jha
ICML 2019 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In our numerical experiments we observe that the GNI formulation always converges to the first-order stationary point of each player s optimization problem. |
| Researcher Affiliation | Industry | 1All authors are with Mitsubishi Electric Research Labs (MERL), Cambridge, MA. |
| Pseudocode | No | The paper describes algorithms and derivations in prose and mathematical equations but does not include a distinct block or figure labeled 'Pseudocode' or 'Algorithm'. |
| Open Source Code | No | The paper does not provide any explicit statement about releasing source code or a link to a code repository for the methodology described. |
| Open Datasets | No | The paper primarily uses simulated data based on specified distributions (e.g., 'Dirac delta distribution', 'Pr = N(µ, I)'). It does not mention any external publicly available datasets with a link, DOI, or formal citation for access. |
| Dataset Splits | No | The paper discusses various game settings (e.g., bilinear two-player games, quadratic games, GANs) and how players are initialized (e.g., 'initialized randomly', 'initialized from N(0, I)'), but it does not provide specific training, validation, or test dataset splits or cross-validation details for its experiments. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., GPU model, CPU model, memory) used for running its experiments. |
| Software Dependencies | No | The paper does not specify any software dependencies with version numbers that would be needed to reproduce the experiments. |
| Experiment Setup | Yes | For GNI, we use a step-size η = 1/L, where L = Q , and ρ = 0.01, while for other methods we use a step-size of η = 0.0011. The methods are initialized randomly the initialization is seen to have little impact on the convergence of GNI, however changed drastically for that of others. We used L = 2, η = ρ = 1/L and initialized all players uniformly from [0, 4]. |