Computing Quantal Stackelberg Equilibrium in Extensive-Form Games

Authors: Jakub Černý, Viliam Lisý, Branislav Bošanský, Bo An5260-5268

AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We experimentally demonstrate that our algorithm provides higher quality results several orders of magnitude faster than a baseline method for general non-linear optimization.
Researcher Affiliation Academia 1 Nanyang Technological University, Singapore 2 AI Center, FEE, Czech Technical University in Prague, Czech Republic
Pseudocode Yes Algorithm 1: Dinkelbach-Type Algorithm for QSE
Open Source Code No The paper cites an appendix for full proofs and additional examples ('Cerny et al. 2021. Computing Quantal Stackelberg Equilibrium in Extensive Form Games: Appendix. https://cloud.disroot.org/s/ 4Cin5Ny3nm Zz Wk R. Accessed: 2021-03-15.'), but this link points to supplementary paper material (PDFs), not source code for the methodology.
Open Datasets No The paper uses 'Search Game' and 'Network Game' domains where instances are constructed based on described rules and random parameters, rather than utilizing pre-existing, publicly available datasets with concrete access information (links, DOIs, or formal citations).
Dataset Splits No The paper describes how game instances are generated and specifies algorithm parameters ('The tolerance parameter for the COBYLA algorithm in NLOPT was set to 10 2 and ϵB = 1% of the leader s utility range for the DTA s binary search. The linearization uses K = 3, the basis of MDT is set to b = 3 and the size of the precision interval E is L = 4.'), but it does not specify train/validation/test splits for any dataset, as the experiments involve generated game instances.
Hardware Specification Yes The experiments were performed on a 3.2GHz CPU with 16GB RAM.
Software Dependencies Yes All implementations were done in C++17. We used NLOPT 2.6.1, and a single-threaded IBM CPLEX 12.8 carried all MILP computations.
Experiment Setup Yes The tolerance parameter for the COBYLA algorithm in NLOPT was set to 10 2 and ϵB = 1% of the leader s utility range for the DTA s binary search. The linearization uses K = 3, the basis of MDT is set to b = 3 and the size of the precision interval E is L = 4.