Community Recovery with Variational Inference and Stochastic Block Models

No Thumbnail Available

URL

Journal Title

Journal ISSN

Volume Title

Perustieteiden korkeakoulu | Master's thesis

Date

2024-05-20

Department

Major/Subject

Machine Learning, Data Science and Artificial Intelligence

Mcode

SCI3044

Degree programme

Master’s Programme in Computer, Communication and Information Sciences

Language

en

Pages

36

Series

Abstract

Many types of data are generated by a process that depends on some grouping or community structure of the elements. In a lot of cases, one is interested in obtaining this community structure using the observed data. This thesis examines the use of variational inference methods to estimate the community structure, modelling the data generation process using stochastic block models. Included is the review of two algorithms, the assumptions they are based on, and theoretical properties they exhibit. In addition, a generalization of one of the algorithms is also derived. Finally, numerical experiments are performed to characterize the accuracy obtained in the community recovery problem. The results show that even with random initializations, these algorithms perform well and reasonably quickly, making them a good choice for real-life problems.

Description

Supervisor

Leskelä, Lasse

Thesis advisor

Leskelä, Lasse

Keywords

Community recovery, Stochastic Block Model, Variational Inference, Random Graphs

Other note

Citation