The Equivalence of Dynamic and Strategic Stability under Regularized Learning in Games
Authors: Victor Boone, Panayotis Mertikopoulos
NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In all our experiments, we ran the EXP3 variant of bandit FTRL (B-FTRL) (cf. Algorithm 3) with step-size and sampling radius parameters πΎπ‘= 0.2 π‘ 1/2 and πΏπ‘= 0.1 π‘ 0.15 respectively. The algorithm was run for π= 104 iterations and, to reduce graphical clutter, we plotted only every third point of each trajectory. Trajectories have been colored throughout with darker hues indicating later times (e.g., light blue indicates that the trajectory is closer in time to its starting point, darker shades of blue indicate proximity to the termination time). The algorithm s initial conditions were taken from a uniform initialization grid of the form π¦1 { 1, 0, 1}3 and perturbed by a uniform random number in [ 0.1, 0.1] to avoid non-generic initializations. |
| Researcher Affiliation | Academia | Victor Boone Panayotis Mertikopoulos Univ. Grenoble Alpes, CNRS, Inria, Grenoble INP, LIG, 38000 Grenoble, France {victor.boone,panayotis.mertikopoulos}@univ-grenoble-alpes.fr |
| Pseudocode | No | The paper describes algorithms using structured equations and textual descriptions (e.g., (RL), (HEDGE), (FTRL)), but these are not presented in explicitly labeled 'Algorithm' or 'Pseudocode' blocks. |
| Open Source Code | No | The paper does not provide any explicit statement about releasing code or a link to a code repository for the methodology described. |
| Open Datasets | No | The numerical experiments section describes using specific "2x2x2 games" with payoff tables, which appear to be custom-defined game settings for their experiments, without mentioning or providing access information for any publicly available or open dataset. |
| Dataset Splits | No | The paper describes running algorithms on custom-defined game scenarios but does not provide specific details on training, validation, or test dataset splits or any methodology for partitioning data for reproduction. |
| Hardware Specification | No | The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory, or processor types) used for running the experiments. |
| Software Dependencies | No | The paper mentions algorithm names like 'EXP3 variant of bandit FTRL (B-FTRL)' and 'Python' for implementation, but it does not specify version numbers for any software, libraries, or dependencies used. |
| Experiment Setup | Yes | In all our experiments, we ran the EXP3 variant of bandit FTRL (B-FTRL) (cf. Algorithm 3) with step-size and sampling radius parameters πΎπ‘= 0.2 π‘ 1/2 and πΏπ‘= 0.1 π‘ 0.15 respectively. The algorithm was run for π= 104 iterations and, to reduce graphical clutter, we plotted only every third point of each trajectory. The algorithm s initial conditions were taken from a uniform initialization grid of the form π¦1 { 1, 0, 1}3 and perturbed by a uniform random number in [ 0.1, 0.1] to avoid non-generic initializations. |