The PhD defence will be in Abels utsikt - Niels Henrik Abels hus. The host of the session will moderate the technicalities while the chair of the defence will moderate the disputation.
Ex auditorio questions: the chair of the defence will invite the audience to ask questions ex auditorio at the end of the defence.
- Submit request to access (available from 5th of June 1:15 pm until 19th of June 1:15 pm)
Trial lecture
19th of June, time: 10:15 am, room 1259 (Abels utsikt).
Auto-regressive shape generation
Main research findings
The profound impact of machine learning is increasingly evident as new tools are adopted by technical communities and the public. With the current pace of scientific advancement, machine learning is poised to further amplify its impact on many facets of society.
This paradigm is equally transformative in geometry. Unlike other disciplines, where success is attributed to scaling to larger datasets and computational resources, machine learning in geometry relies on improving the algorithmic aspects. Past research has improved understanding, representation power, speed, and the capacity to handle new input and output modalities, resulting in applications like object synthesis and scene reconstruction. However, challenges persist in model selection, understanding limitations, internal mechanisms, and ensuring guarantees and control.
This thesis explores the interplay of machine learning and geometry, applying tools from classical geometry and topology to modern neural networks to develop new perspectives and computational tools for understanding, interpreting, manipulating, and generating neural geometry representations. It culminates in training a generative model without data, relying solely on available theory. This distinct approach underscores a novel paradigm in machine learning. The tools introduced in this thesis not only improve current methods but also pave the way for entirely novel research questions.
Adjudication committee
- Professor Panagiotis Kaklis, University of Strathclyde
- Reader Amir Vaxman, University of Edinburgh
- Associate Professor Håkon Hoel, University of Oslo
Supervisors
- Research Scientist Georg Muntingh, SINTEF
- Professor Michael S. Floater, University of Oslo
- Research Manager Tor Dokken, SINTEF
- Research Scientist Oliver Barrowclough, SINTEF
Chair of defence
Head of Department Nils Henrik Risebro.
Host of the session
Associate Professor Håkon Hoel, University of Oslo