Decentralized Planning in Stochastic Environments with Submodular Rewards
Authors: Rajiv Kumar, Pradeep Varakantham, Akshat Kumar
AAAI 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | Finally, we provide experimental results on two benchmarks from the Dec-MDP literature, where we show that the posteriori quality guarantees are significantly better (in some cases close to 90% of optimal) than the a priori guarantee of at least 50% from optimal. We focus primarily on the cooperative case with TI-Dec MDPs to verify the online guarantees provided by our approaches. 5.1 Security Games For this domain, we experiment with the problem domain provided by Shieh et al. (Shieh et al. 2014). |
| Researcher Affiliation | Academia | Rajiv Ranjan Kumar, Pradeep Varakantham, Akshat Kumar School of Information Systems Singapore Management University {rajivk,pradeepv,akshatkumar}@smu.edu.sg |
| Pseudocode | Yes | Algorithm 1: GREEDY |
| Open Source Code | No | The paper does not contain any explicit statements or links indicating that the source code for the described methodology is publicly available. |
| Open Datasets | No | The paper mentions using 'benchmarks from the Dec-MDP literature' and specific problem domains provided by cited works (Shieh et al. 2014; Nair et al. 2005; Kumar and Zilberstein 2011). However, it does not provide direct URLs, DOIs, specific repository names, or explicit statements about where to access the exact datasets or problem instances used for the experiments. It describes parameters varied for the experiments but not how to obtain the underlying data for reproduction. |
| Dataset Splits | No | The paper describes varying parameters for its simulation-based experiments (e.g., 'Number of targets = {10, 12, 14, 16, 18, 20}', 'Number of agents was varied from 10-60'), but it does not specify any training, validation, or test dataset splits in terms of percentages, sample counts, or references to predefined splits typically used for model training and evaluation on datasets. |
| Hardware Specification | No | The paper does not provide any specific details about the hardware used to run the experiments, such as CPU or GPU models, memory specifications, or cloud computing instance types. |
| Software Dependencies | No | The paper describes the algorithms and their application to problem domains but does not list any specific software dependencies (e.g., programming languages, libraries, frameworks, or solvers) along with their version numbers that would be required to reproduce the experiments. |
| Experiment Setup | Yes | We experimented with the following values of the different parameters: (i) Number of targets = {10, 12, 14, 16, 18, 20}; (ii) Number of agents was varied from 10-60, depending on the number of targets; (iii) Effectiveness parameter, ϵ = {0.3, 0.5, 0.7}. We primarily experimented with lazy greedy, as greedy was unable to solve problems beyond the smallest ones and lazy greedy is more efficient without losing on solution quality. |