Disputas - Master of Science Sigurd Løkse

"Leveraging Kernels for Unsupervised Learning"

 

Popular scientific abstract:

Within the field of machine learning, unsupervised learning refers to learning with no (or minimal) support from labels or prior knowledge. Due to large amounts of unlabeled data being available, and manually labelling data being resource intensive, unsupervised learning will be increasingly important in the future for machine learning, with applications in e.g. medical data, marketing, logistics and identifying fraudulent activity. To this end, we address several challenges within unsupervised learning, and develop novel methodology by leveraging the so-called kernels and kernel methods. The contributions of this thesis are two-fold.

Firstly, we focus on pure kernel methods and develop a kernel for data with missing elements which automatically adapts to the inherent structures in the data. In addition we develop new kernel-based ranking methodology. Secondly, we propose several novel deep learning methods which are fused with kernels in order to solve unsupervised learning tasks.

The thesis is published in Munin and is available at: https://hdl.handle.net/10037/19911

 

Supervisors:

• Professor Robert Jenssen, Department of Physics and Technology, UiT (main supervisor)
• Senior Research Scientist Arnt-Børre Salberg, Norwegian Computing Center

 

 

Evaluation committee:

• Professor Gustau Camps-Valls, Department of Electrical Engineering, University of Valencia, Spain (1. opponent)
• Senior Lecturer Sarah Filippi, Department of Mathematics, Imperial College London, UK (2. opponent)
• Associate Professor Stian Normann Anfinsen, Department of Physics and Technology, UiT (internal member and leader of the committee)

Both opponents will participate remotely to the defence.

 

Leader of the public defense:
The leader of the public defense is Professor Arne O. Smalås, Dean, Faculty of Science and Technology, UiT.

 

Opposition ex auditorio:
If you have any questions for the candidate during the public defence, please send an e-mail to the leader of the public defence. They will announce the questions during the defence.

 

Trial lecture:

The trial lecture is held Friday December 11th at 10:15 AM in the same auditorium.

Title of the trial lecture: «Bayesian Deep Learning»

 

Streaming:

The defense and trial lecture will be streamed via Mediasite: 

https://mediasite.uit.no/Mediasite/Catalog/Full/1a2f904b8f5542ca8d829da14a4e6fc821

 

Audience:

UiT follows the national guidelines regarding infection control. A maximum of 20 people are allowed in the auditorium during the defence, as long as everybody keeps a distance of 1 meter at all times.