Public defence in Computer Science, M.Sc. Paul Chang

Title of the thesis: Rethinking Inference in Gaussian Processes: A Dual Parameterization Approach

Doctoral student: Paul Chang
Opponent: Professorr Andrew Gordon Wilson, New York University, USA
Custos: Assistant Professor Arno Solin, Aalto University School of Science, Department of Computer Science

Artificial intelligence is rapidly advancing in applications ranging from chat assistants to self-driving cars. At the core of these technologies lies the need for accurate predictions. But how certain are we about these predictions? For example, a bus with an ETA of 5 minutes and a 10-minute range is quite different from the exact ETA with a 1-minute range. The field of uncertainty quantification aims to measure confidence in predictions and has applications across many domains. Machine learning models that provide uncertainty quantification are crucial for fields like climate science, medicine, and finance, especially when decision-making based on model predictions has critical implications.

This thesis improves the applicability of Gaussian Processes (GPs) as a model class for uncertainty quantification. It attempts to extend uncertainty quantification in environments with non-Gaussian data and potentially large data sets. This advancement enables better GP performance in areas like continual learning (learning from data streams over time) and Bayesian optimisation, where efficiency and adaptability are essential. By enhancing GPs' capacity to handle complex data and quantify uncertainty effectively, this research supports broader machine learning applications, helping ensure reliable forecasting and decision-making in uncertain environments.

Key words: Machine learning, Uncertainty Quantification, Sequential Decision Making

Thesis available for public display 10 days prior to the defence at: https://aaltodoc.aalto.fi/doc_public/eonly/riiputus/ 

Contact information:

Doctoral theses of the School of Science: https://aaltodoc.aalto.fi/handle/123456789/52 

Zoom Quick Guide: https://www.aalto.fi/en/services/zoom-quick-guide