Group Testing on a Network
Authors: Arlei Silva, Ambuj Singh4348-4356
AAAI 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We evaluate our algorithms and baselines using epidemic processes on real and synthetic contact networks. |
| Researcher Affiliation | Academia | Arlei Silva, Ambuj Singh University of California, Santa Barbara |
| Pseudocode | Yes | Algorithm 1 Dorfman's Group Test; Algorithm 2 Greedy; Algorithm 3 Kernighan-Lin |
| Open Source Code | Yes | Code: https://github.com/arleilps/group-testing |
| Open Datasets | Yes | Primary School (PS), High School (HS), Company (CP), and Conference (CF) are real face-to-face contact networks over varying periods of time (from 2 days to 2 weeks) from sociopatterns (G enois and Barrat 2018). Erdos-Renyi (ER) and Gaussian Random Partition (GRP) are unweighted synthetic graphs generated with the respective models (Erd os and R enyi 1959; Brandes, Gaertler, and Wagner 2003). Gowalla (GW) is a co-location network based on user check-ins from the (now extinct) Gowalla social network (Liu et al. 2013). |
| Dataset Splits | No | The paper describes running simulations on entire networks and partitioning nodes into groups, but it does not specify explicit training, validation, and test dataset splits in terms of percentages or sample counts for model training or evaluation in the conventional machine learning sense. |
| Hardware Specification | No | The paper mentions running times for experiments ('For 800 vertices, KL-Sampling and KL-Topology take approximately 600 and 200 secs to finish, respectively.'), but it does not provide specific details about the hardware used, such as CPU/GPU models or memory. |
| Software Dependencies | No | The paper states, 'Our simulations are implemented using the open-source EON Python module', but it does not provide specific version numbers for Python or the EON module. |
| Experiment Setup | Yes | The other parameters of the process are set for each network, according to their weight distribution, to produce a slow progression of the infection, with γ varying from .1 and 1.5 and τ varying from 1 to 40. Each simulation is stopped when q.|V | nodes in the network are infected or recovered, where q [0, 1] is the prevalence value. The number of samples, z, for the sampling-based methods was set to 1,000 in all experiments. |