Policy Learning for Continuous Space Security Games Using Neural Networks

Authors: Nitin Kamra, Umang Gupta, Fei Fang, Yan Liu, Milind Tambe

AAAI 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate the potential to predict good defender strategies via experiments and analysis of Opt Grad FP and Opt Grad FP-NN on discrete and continuous game settings.
Researcher Affiliation Academia University of Southern California1, Carnegie Mellon University2 {nkamra, umanggup, yanliu.cs, tambe}@usc.edu1, feifang@cmu.edu2
Pseudocode Yes Algorithm 1: Opt Grad FP
Open Source Code No The paper does not contain any explicit statement about releasing source code for the described methodology or a link to a code repository.
Open Datasets No The paper describes using "1000 randomly generated forest states" for training but does not provide any specific access information (link, DOI, repository, or citation) to make this dataset publicly available or reproducible.
Dataset Splits No The paper mentions training on "1000 randomly generated forest states" and testing on "10 new forest states", indicating a train/test split, but it does not specify a validation set or precise percentage/counts for these splits, nor does it refer to predefined splits for these custom generated states.
Hardware Specification No The paper does not specify the exact hardware used for running experiments, such as CPU or GPU models, memory, or cloud instance types.
Software Dependencies No The paper mentions software components like convolutional neural networks and policy gradients, but it does not provide specific version numbers for any programming languages, libraries, or frameworks (e.g., Python, TensorFlow, PyTorch, CUDA) used in the implementation.
Experiment Setup Yes The forest game s hyperparameters for the single forest state case are summarized in Table 4. Opt Grad FP-NN for multiple forest states uses the same parameters except epmax = 20000 and E = 500000. The architectures of all neural networks presented earlier and all algorithm hyperparameters were chosen by doing informal grid searches within appropriate intervals.