Title:	Machine Learning for Scientific Discovery
DNr:	Berzelius-2026-23
Project Type:	LiU Berzelius
Principal Investigator:	Hossein Azizpour <azizpour@kth.se>
Affiliation:	Kungliga Tekniska högskolan
Duration:	2026-02-01 – 2026-08-01
Classification:	20208
Homepage:	https://www.csc.kth.se/~azizpour/
Keywords:

Abstract

This proposal is for the scientific studies within the group of Hossein Azizpour. In Azizpour's group we try to understand modern deep networks and train them for impactful applications. Therefore, there are two main types of projects in the group. One is on the fundamentals of deep learning and the other on the application side. Fundamental of deep networks: here three main tracks are pursued for the foundations of trustworthy deep network: (i) generative modeling with diffusion and flow matching models, (ii) interpreting and understanding trained deep networks especially from the lens of a functional analysis, (iii) explaining individual decisions of a trained deep network in a human-interpretable way, and (iv) quantification and robustness to uncertainty in the decisions of a trained deep network. Applications: on the application side, we apply the findings of the fundamental research on a few impactful applications including: (i) general computer vision, (ii) earth observation, especially uncertainty in the developments for detecting urbanization and forest fires, (iii) protein structure modelling, especially the interpretable models and uncertainty of predictions, and (iv) modeling air flow in simulations and from real-world measurements using deep generative models. As such, all of the directions are either purely empirical or require empirical validations of the theories and long-trained generative models. Such empirical investigations for modern deep networks such as large ResNets and visual transformers can only be enabled with the help of a large GPU cluster such as Berzelius. This proposal is for continuation of all the ongoing projects within Azizpour's group. The currently active project Berzelius-2025-339 has been used up and is requested to be merged with an additional project for its remaining lifetime (2 months) and additional 4 months for 16000 GPU-hours per month.