Evolutionary Approach to Security Games with Signaling
Authors: Adam Żychowski, Jacek Mańdziuk, Elizabeth Bondi, Aravind Venugopal, Milind Tambe, Balaraman Ravindran
IJCAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | In the majority of 342 test game instances, EASGS outperforms state-of-the-art methods, including a reinforcement learning method, in terms of time scalability, nearly constant memory utilization, and quality of the returned defender s strategies (expected payoffs). |
| Researcher Affiliation | Academia | 1Faculty of Mathematics and Information Science, Warsaw University of Technology 2Center for Research on Computation and Society, Harvard University 3 Robert Bosch Centre for Data Science and AI, IIT Madras 4Department of Computer Science and Engineering, IIT Madras |
| Pseudocode | No | The paper describes the algorithm's steps and logic in prose and mathematical notation but does not include structured pseudocode or a formally labeled algorithm block. |
| Open Source Code | Yes | EASGS source code can be found on github.com/easgs/source code. |
| Open Datasets | Yes | All generated games are publicly available on github.com/easgs/benchmark games. |
| Dataset Splits | Yes | EASGS parameters were tuned on a set of 12 games with 20 vertices (3 games of each type: sparse, moderate, dense and locally-dense), which were separated from the 342 EASGS benchmark graphs and not used during the method evaluation. |
| Hardware Specification | Yes | Tests were performed on a cluster running Cent OS Linux 7 (Core) with Intel(R) Xeon(R) CPU E5-2683 v4 @ 2.1 GHz with 128 GB RAM and 4 cores. |
| Software Dependencies | No | The paper mentions 'Cent OS Linux 7 (Core)' for the operating system but does not provide specific version numbers for programming languages, libraries, or other key software dependencies (e.g., Python, PyTorch, TensorFlow versions) used in the experiments. |
| Experiment Setup | Yes | Based on 5000 runs, the following parameter values were finally chosen: population size npop = 200, crossover probability Pc = 0.5, mutation probability Pm = 0.8, mutation repetition limit mlimit = 10, number of elite chromosomes ne = 2, selection pressure Psp = 0.8, generations limit ngen = 2000, number of generations between refreshes nref = 300. |