Choices Are Not Independent: Stackelberg Security Games with Nested Quantal Response Models
Authors: Tien Mai, Arunesh Sinha5141-5149
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we conduct experiments that show the scalability and parallelizability of our approach, as well as advantages of the nested QR model. Our results show the practical scalability of our algorithm and the ability to exploit parallel computation, and compares it with other baselines as well as other possible design choices in our algorithm. We also show in simulation that the nested QR model better models human response when there are a large number of correlated choices, thereby, providing higher utility to the defender in a SSG. |
| Researcher Affiliation | Academia | School of Computing and Information Systems, Singapore Management University atmai@smu.edu.sg, aruneshs@smu.edu.sg |
| Pseudocode | Yes | Algorithm 1: Subproblemn(Mn); Algorithm 2: Bin Search Dyn Prog(M) |
| Open Source Code | No | The paper does not contain any explicit statement about releasing the source code for the described methodology, nor does it provide a link to a code repository. |
| Open Datasets | No | The paper states, "We generate 8 random instances for each measurement that we report." This indicates the use of randomly generated instances rather than a publicly available dataset. No specific links, DOIs, repositories, or citations for a public dataset are provided. |
| Dataset Splits | No | The paper mentions generating "random instances" but does not specify any training, validation, or test dataset splits (e.g., percentages, sample counts, or references to predefined splits). |
| Hardware Specification | Yes | All the experiments were conducted using Matlab on a PC with Intel(R) Core(TM) i7-9700 CPUs running at 3.00GHz. We use a single CPU core for all the experiments except the computational scalability part where 8 CPU cores are used to demonstrate how our algorithm scales up with parallel computing. |
| Software Dependencies | No | The paper states that experiments were conducted using "Matlab" but does not provide a specific version number for Matlab or any other ancillary software components used. |
| Experiment Setup | Yes | For (sub OPT-2), in order to have good approximations, we simply select ρ such that the number of grid points is 10. We then pick the interval between two consecutive points that achieves the best values and generate another set of 5 points to further improve the solution within this interval. Overall, this only requires to compute g(Wn) 15 times... For the discretized dynamic program (DYN-APPROX)... we vary the number of grid points (denoted as T) from 20 to 120... Based on this observation, we use T = 100 for rest of the experiments. We generate 8 random instances for each measurement that we report. For each game instance, each target is assigned randomly to a nest Kn, n [N], and the nest parameters σn are generated randomly in [0.5, 1]. We also choose the resource parameter as M = P n [N] |Kn|/5. |