Near-Optimal $Φ$-Regret Learning in Extensive-Form Games
Authors: Ioannis Anagnostides, Gabriele Farina, Tuomas Sandholm
ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we support our theory (Theorem 1.1) by implementing our algorithm and evaluating its performance through experiments on several benchmark extensive-form games in Section 4. |
| Researcher Affiliation | Collaboration | 1Department of computer science, Carnegie Mellon University, Pittsburgh, USA 2FAIR, Meta AI 3Strategy Robot, Inc. 4Optimized Markets, Inc. 5Strategic Machine, Inc. |
| Pseudocode | Yes | Algorithm 1 Φ-Regret Minimizer (Gordon et al., 2008) |
| Open Source Code | No | No statement regarding open-source code availability or repository links for the described methodology was found. |
| Open Datasets | Yes | Finally, in this section we experimentally verify our theoretical results on several common benchmark extensive-form games: (i) 3-player Kuhn poker (Kuhn, 1953); (ii) 2-player Goofspiel (Ross, 1971); and (iii) 2-player Sheriff (Farina et al., 2019c). |
| Dataset Splits | No | The paper describes experiments on game theory scenarios (Kuhn poker, Goofspiel, Sheriff) which involve repeated play, but it does not specify traditional dataset splits (e.g., percentages or counts for training, validation, and test sets) as commonly found in supervised learning. |
| Hardware Specification | No | No specific hardware details (e.g., GPU/CPU models, memory) used for running the experiments were provided in the paper. |
| Software Dependencies | No | The paper names algorithms but does not provide specific version numbers for any software dependencies or libraries. |
| Experiment Setup | Yes | For simplicity we use the same learning rate η > 0 for all the local regret minimizers, which is treated as a hyperparameter in order to obtain better empirical performance. In particular, after a very mild tuning process, we chose η = 1 for all our experiments. |