Dynamic Network Discovery via Infection Tracing

Authors: Ben Bals, Michelle Döring, Nicolas Klodt, George Skretas

IJCAI 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We round off our analysis with an experimental evaluation of our algorithm on real-world interaction data from the Stanford Network Analysis Project and on temporal Erd os-Renyi graphs.In Section 7, we empirically validate our theoretical results. Using both synthetic and real-world data, we execute the Discovery Follow algorithm and observe its performance.
Researcher Affiliation Academia Ben Bals1,2 , Michelle D oring3 , Nicolas Klodt3 and George Skretas3 1Centrum Wiskunde & Informatica, Amsterdam 2Vrije Universiteit Amsterdam 3Hasso Plattner Institute, Potsdam EMAIL, EMAIL,
Pseudocode Yes Algorithm 1: Follow discovers the neighbors of v0. Explore discovers their respective δ-ecc s. and Algorithm 2: The TGD extension of Follow.
Open Source Code Yes Both our implementation and analysis code are available under a permissive open-source license and can be used to replicate our findings.1 1See https://github.com/Ben Bals/dynamic-network-discovery.
Open Datasets Yes Secondly, to evaluate our algorithm on real-world data, we employ a data set from the Stanford Large Network Dataset Collection [Kumar et al., 2021]. The comm-f2f-Resistance collection is described by the project as a set of 62 dynamic face-to-face interaction network[s] between a group of participants .
Dataset Splits No No explicit dataset splits (training/test/validation) are mentioned. The paper describes generating synthetic temporal Erd os-Renyi graphs with parameters n, p, Tmax and using the entire comm-f2f-Resistance dataset for evaluation, rather than splitting it for typical machine learning tasks.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts) used for running the experiments. It only states 'We run Discovery Follow on all graphs...'.
Software Dependencies No The paper states that implementation and analysis code are open-source but does not list specific software dependencies with version numbers (e.g., programming languages, libraries, or frameworks with their versions).
Experiment Setup Yes We test with 1 to 100 nodes in steps of size 5, for probabilities p {.01, .05, .1, .15, .2, .25, .3, .35, .4, .5, .7, .9}. We pick Tmax as a factor of n, testing with Tmax/n {.05, .1, .2, .3, .4, .5, .7, .9, 1, 2, 3, 5, 7, 10}. Similarly, δ is 1 or a multiple of Tmax, namely δ/Tmax {.01, .05, .1, .3, .5} We record the number of rounds needed to complete the discovery and how many of these rounds are spent in the component discovery versus the component exploration phases of the algorithm.