Uncertainty estimation for machine learning system self-assessment on medical conversations
Jakob Havtorn (Ph.D. Student)
This project will develop extensions of probabilistic generative modelling to efficiently handle high-dimensional sequence and time-series data, such as audio, and show that robust semi-supervised learning, reliable out-of-distribution detection and uncertainty estimation can be achieved. In real-world applications, well-calibrated uncertainty estimates translate into improved robustness and allows machine learning systems to enter into reliable service in a wide range of critical applications from medical triaging and radiology over autonomous vehicles to financial services and transaction validation. Uncertainty awareness will ultimately help make more efficient use of resources in critical decision-making and make it clear when human intervention is required.
|Primary Advisor:||Jes Frellsen (Technical University of Denmark)|
|Industry Advisor:||Lars Maaløe (Corti.ai & Technical University of Denmark)|
|PhD Duration:||01 September 2020 - 31 August 2023|