Skip to content Search
Search our website:

Graph generative networks for hierarchical ontologies

Graph generative networks for hierarchical ontologies

Principal supervisor: Dr Cen Wan and Dr Carl Barton

The recent success of graph neural networks and network embedding methods enables us to learn high-level feature representations of graph-structured data. With the help of generative networks (e.g. generative adversarial networks), the original graph structure can also be reconstructed based on informative latent spaces. The hierarchical ontology is a well-known semantic-rich knowledge representation paradigm, where vertices are organised as a directed acyclic graph by hierarchical relationships. However, the structure of hierarchical ontologies is usually complex, leading to a challenge for existing graph generation methods. This project is going to investigate and develop novel machine learning algorithms that are able to cope with the task of hierarchical ontology regeneration. The main research theme includes developing novel graph generative networks, supervised/unsupervised hierarchical ontology optimisation, etc. In addition, the developed methods will be applied to analysing biological data, e.g. the well-known Gene Ontology.

Candidate Requirements:

This PhD is aimed at graduates with good machine learning background knowledge on deep neural networks and graph theory as well as excellent programming skills.

Key References:

[1] A. Bojchevski, O. Shchur, D. Zugner, S. Gunnemann. NetGAN: Generating Graphs via Random Walks, ICML 2018.
[2] I. Goodfellow, J. Pouget-Abadie, M. Mirza, B. Xu, D. Warde-Farley, S. Ozair, A. Courville, Y. Bengio. Generative Adversarial Nets, NIPS 2014.
[3] J. You, R. Ying, X. Ren, W.L. Hamilton, J. Leskovec. GraphRNN: Generating Realistic Graphs with Deep Auto-regressive Models, ICML 2018.

Further details about the project may be obtained from:

Principal supervisor: Dr Cen Wan (Enable JavaScript to view protected content.) and Dr Carl Barton (Enable JavaScript to view protected content.)

Further information about PhDs at CSIS is available via this link.