Gu et al., "Principled approach to the selection of the embedding dimension of networks" (2021)

2021-06-18 → 2026-08-14

Weiwei Gu, Aditya Tandon, Yong-Yeol Ahn, and Filippo Radicchi, Nature Communications 12, 3772 (2021)
Link | arXiv | PDF | Code

@article{gu2021defining,
    author = {Weiwei Gu and Aditya Tandon and Yong-Yeol Ahn and Filippo Radicchi},
    title = {Principled approach to the selection of the embedding dimension of networks},
    journal = {Nature Communications},
    volume = {12},
    number = {1},
    pages = {3772},
    doi = {10.1038/s41467-021-23795-5},
    archivePrefix = {arXiv},
    eprint = {2004.09928},
    primaryClass = {physics.soc-ph},
    year = {2021},
}

Proposes a method for choosing a network embedding dimension that parsimoniously preserves structural information rather than maximizing performance on a downstream task. Tests across multiple embedding algorithms and real networks suggest that low-dimensional encodings are often sufficient.

Graph embedding, Dimensionality reduction, Model selection

Receive my updates