Privacy-preserving and Explainable Machine Learning for Healthcare
Date and time: 2 May 2023, 13:00 – 14:00 CEST (UTC +2)
Speaker: Jens Lundström, Senior Lecturer at Halmstad University (HU)
Title: Privacy-preserving and Explainable Machine Learning for Healthcare
Where: Digital Futures hub, Osquars Backe 5, floor 2 at KTH main campus OR Zoom
Administrator: Rami Mochaourab, firstname.lastname@example.org
Abstract: Machine learning-based models have been deployed and used for years in various industries. As these models become more complex and integrated into daily decisions, several needs arise. This talk will focus on two questions related to those needs; 1) how to preserve privacy during model training and after deployment, and 2) how to explain specific models such as Graph Neural Networks. This talk will introduce the concepts of privacy-preserving machine learning, with a specific focus on synthetic data generation as well as federated learning methods. Ongoing research and feasibility studies at Halmstad University will be highlighted.
Bio: Jens Lundström is a Senior Lecturer at Halmstad University (HU) Sweden and an engineer specializing in artificial intelligence. His duties at HU include research, leadership and teaching in the domain of Machine Learning and AI. His interests are in applied and theoretical ML research for improving healthcare, patient experience and quality of life. Jens Lundström’s recent work is on the explainability of machine learning-based systems, synthetic data generation, federated learning, and risk mitigation for model attacks.
Jens graduated with a licentiate degree in Information Technology from Örebro University in 2012 and received his PhD in Information Technology from Halmstad University in 2014.
Besides teaching and research at HU, he works as a research area leader for AI in the research profile CAISR Health, as well as a coordinator for Halmstad Professionals, a network for individuals interested in AI, data analytics, cyber security, service design and business intelligence.