Deep Learning Attention Model for Supervised and Unsupervised Network Community Detection

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F00216224%3A14310%2F23%3A00137485" target="_blank" >RIV/00216224:14310/23:00137485 - isvavai.cz</a>

Výsledek na webu

<a href="https://link.springer.com/chapter/10.1007/978-3-031-36027-5_51" target="_blank" >https://link.springer.com/chapter/10.1007/978-3-031-36027-5_51</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-031-36027-5_51" target="_blank" >10.1007/978-3-031-36027-5_51</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Deep Learning Attention Model for Supervised and Unsupervised Network Community Detection

Popis výsledku v původním jazyce

Network community detection is a complex problem that has to utilize heuristic approaches. It often relies on optimizing partition quality functions, such as modularity, description length, stochastic block-model likelihood etc. However, direct application of the traditional optimization methods has limited efficiency in finding the global maxima in such tasks. This paper proposes a novel bi-partite attention graph neural network model for supervised and unsupervised network community detection, suitable for unsupervised optimization of arbitrary partition quality functions, as well as for minimization of a loss function against the provided partition in a supervised setting. The model is demonstrated to be helpful in the unsupervised improvement of suboptimal partitions previously obtained by other known methods like Louvain algorithm for some of the classic and synthetic networks. It is also shown to be efficient in supervised learning of the provided community structure for a set of classic and synthetic networks. Furthermore, the paper serves as a proof-of-concept for the broader application of graph neural network models to unsupervised network optimization.

Název v anglickém jazyce

Deep Learning Attention Model for Supervised and Unsupervised Network Community Detection

Popis výsledku anglicky

Network community detection is a complex problem that has to utilize heuristic approaches. It often relies on optimizing partition quality functions, such as modularity, description length, stochastic block-model likelihood etc. However, direct application of the traditional optimization methods has limited efficiency in finding the global maxima in such tasks. This paper proposes a novel bi-partite attention graph neural network model for supervised and unsupervised network community detection, suitable for unsupervised optimization of arbitrary partition quality functions, as well as for minimization of a loss function against the provided partition in a supervised setting. The model is demonstrated to be helpful in the unsupervised improvement of suboptimal partitions previously obtained by other known methods like Louvain algorithm for some of the classic and synthetic networks. It is also shown to be efficient in supervised learning of the provided community structure for a set of classic and synthetic networks. Furthermore, the paper serves as a proof-of-concept for the broader application of graph neural network models to unsupervised network optimization.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10100 - Mathematics

Projekt

<a href="/cs/project/EF16_019%2F0000822" target="_blank" >EF16_019/0000822: Centrum excelence pro kyberkriminalitu, kyberbezpečnost a ochranu kritických informačních infrastruktur</a><br>

Návaznosti

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)<br>S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2023

Kód důvěrnosti údajů

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Název statě ve sborníku

Computational Science – ICCS 2023

ISBN

9783031360268

ISSN

0302-9743

e-ISSN

1611-3349

Počet stran výsledku

8

Strana od-do

647-654

Název nakladatele

Springer Cham

Místo vydání

Cham

Místo konání akce

Prague

Datum konání akce

3. 7. 2023

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—