Decentralized Submodular Maximization: Bridging Discrete and Continuous Settings
Authors: Aryan Mokhtari, Hamed Hassani, Amin Karbasi
ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | 7. Numerical Experiments: We will consider a discrete setting for our experiments and use Algorithm 2 to find a decentralized solution. The main objective is to demonstrate how consensus is reached and how the global objective increases depending on the topology of the network and the parameters of the algorithm. For our experiments, we have used the Movie Lens data set. It consists of 1 million ratings (from 1 to 5) by M = 6000 users for p = 4000 movies. |
| Researcher Affiliation | Academia | 1Laboratory for Information and Decision Systems, Massachusetts Institute of Technology 2Department of Electrical and Systems Engineering, University of Pennsylvania 3Department of Electrical Engineering and Computer Science, Yale University. |
| Pseudocode | Yes | Algorithm 1 DCG at node i; Algorithm 2 Discrete DCG at node i |
| Open Source Code | No | The paper does not provide any statements or links indicating that source code for the described methodology is openly available. |
| Open Datasets | No | The paper mentions using the 'Movie Lens data set' but does not provide a specific citation with author names and year, a direct URL, DOI, or repository information for accessing this dataset. |
| Dataset Splits | No | The paper describes how the MovieLens data is partitioned across network nodes but does not specify traditional training, validation, and test splits for model evaluation. |
| Hardware Specification | No | The paper does not provide specific details about the hardware used for running the experiments, such as CPU/GPU models or memory specifications. |
| Software Dependencies | No | The paper does not mention any specific software dependencies with version numbers. |
| Experiment Setup | Yes | For our experiments, we have used the Movie Lens data set. ... We consider a network of n = 100 nodes. ... The data has been distributed equally between the nodes of the network, i.e., the set of users has been partitioned into 100 equally-sized sets and each node in the network has access to only one chunk (partition) of the data. ... We consider three different choices for the underlying communication graph between the 100 nodes: A line graph ..., an Erdos-Renyi random graph (with average degree 5), and a complete graph. The matrix W is chosen as follows ... Finally we let wi,i = 1 Pj2N wi,j. ... We have run Algorithm 2 for T = 50 and T = 1000. ... If Assumptions 1-5 hold and we set = T 1/2 and φ = T 2/3 |