|Topics in Machine Learning|
|Organisatsiooni nimi||Machine Learning|
|Kokkuvõte||We invite strongly motivated and hard-working students at all levels (bachelor, master, doctoral) to contact us if you are interested in writing your thesis about machine learning methods or theory.|
We believe that the key to advancements in machine learning is a deep understanding of why the current methods succeed or fail. Our main goal is to build trustworthy machine learning models that know how well they know. Some of our current research directions are uncertainty quantification, context modelling, and representation learning, particularly for applications in machine perception.
Some topics for 2022/23:
* Analysing epistemic uncertainty in a controlled setting
* Discovering contexts with Multi Exit Models
* Representing uncertainty in regression
* Gradual freezing for better uncertainty calibration
* Exploring out-of-distribution detection using Bayesian VAEs.
* Experimental comparison of cost-sensitive loss functions
* Experimental analyses of semantic segmentation quality (in unexpected scenarios)
+ ask for more
|Lõputöö kaitsmise aasta||2022-2023|
|Suhtlemiskeel(ed)||eesti keel, inglise keel|