Generative Adversarial Equilibrium Solvers
Authors: Denizalp Goktas, David C. Parkes, Ian Gemp, Luke Marris, Georgios Piliouras, Romuald Elie, Guy Lever, Andrea Tacchetti
ICLR 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 5 EXPERIMENTAL RESULTS Arrow-Debreu exchange economies. Our first set of experiments aim to solve CE in Arrow Debreu exchange economies (Arrow & Debreu, 1954). ... We benchmark GAES to tatonnement (Walras, 1896), which is an auction-like algorithm that is guaranteed to converge for CES utilities with |
| Researcher Affiliation | Collaboration | Denizalp Goktas , Brown University denizalp goktas@brown.edu David C. Parkes , Ian Gemp, Luke Marris, Georgios Piliouras, Romuald Elie, Guy Lever, Andrea Tacchetti Google Deep Mind {parkesd, imgemp, marris, parkesd, imgemp, marris, gpil, relie, guylever, atacchet}@google.com Research conducted while the author was an intern at Google Deep Mind. Also, School of Engineering and Applied Sciences, Harvard University. |
| Pseudocode | Yes | Algorithm 1 Stochastic Exploitability Descent Inputs: B, h, f , Th, Tf , wh,(0), wf ,(0) Outputs: (wh,(t), wf ,(t)) Th t=0 1: for t = 0, . . . , Th 1 do 2: Receive batch B(t) S. 3: wh,(t+1) = wh,(t) (t) rwh b (wh,(t), wf ,(t)) 4: for s = 0, . . . , Tf 1 do 5: Receive batch B(s) S. 6: wf = wf + (s) f rwf b (wh,(t), wf ) 7: end for 8: wf ,(t+1) = wf 9: end for 10: Return (wh,(t), wf ,(t)) |
| Open Source Code | No | The paper does not contain any explicit statement about open-sourcing the code for the described methodology or provide a link to a code repository. |
| Open Datasets | No | The paper describes the parameters and utility functions used to sample or generate problem instances for their experiments (e.g., "sampled as a tuple (V, E) ∈ Rn×m × Rn×m for linear, Cobb-Douglas, and Leontief exchange economies"). However, it does not provide concrete access information (link, DOI, citation to an established public dataset, or statement of availability) for these generated instances. |
| Dataset Splits | No | For each of these baselines, we run an extensive grid search over decreasing learning rates during validation (see Appendix G.2). While "validation" is mentioned, no specific details about the size, percentage, or method of splitting the validation set are provided. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware (e.g., GPU/CPU models, memory) used to run the experiments. |
| Software Dependencies | No | The paper mentions several software tools and libraries (e.g., JAX, Num Py, Matplotlib, seaborn, pycddlib, Python) but does not provide specific version numbers for these dependencies, which are necessary for reproducible setup. |
| Experiment Setup | No | The paper states, "We include experiments on normal-form games, as well as all missing additional implementation details and network architecture diagrams in Appendix G." and "For each of these baselines, we run an extensive grid search over decreasing learning rates during validation (see Appendix G.2)." This indicates that experimental setup details, such as specific hyperparameter values, are deferred to the appendix and not present in the main text. |