Pooling Operators in Graph Neural Networks for Human Activity Recognition

Bondar, George (2023) Pooling Operators in Graph Neural Networks for Human Activity Recognition. Bachelor's Thesis, Computing Science.

Preview	Text bCS_2023_BondarG.pdf Download (1MB) | Preview
	Text toestemming.pdf Restricted to Registered users only Download (130kB)

Abstract

The field of Human Activity Recognition (HAR) has wide-ranging applications in areas such as healthcare, industry, and security. Graph Neural Networks (GNNs) have shown promising results in HAR due to their robustness to missing data and transfer learning capabilities. The accuracy of graph classification models depends to a large extent on the pooling operation. As a consequence, the scope of the research will fall back upon empirically analysing different pooling strategies based on a carefully selected model. The research questions are outlined, and the methodology is presented. The paper also includes background information on both GNNs for HAR and pooling operators. Finally, the paper presents limitations of graph pooling methods and proposes new metrics beyond empirical experiments for evaluating the performance gains of graph pooling models.

Item Type:	Thesis (Bachelor's Thesis)
Supervisor name:	Lazovik, A. and Tello Guerrero, M.A.
Degree programme:	Computing Science
Thesis type:	Bachelor's Thesis
Language:	English
Date Deposited:	20 Nov 2023 10:56
Last Modified:	20 Nov 2023 10:56
URI:	https://fse.studenttheses.ub.rug.nl/id/eprint/31645

Actions (login required)

View Item