Thema der Dissertation:
Machine learning-driven data integration for drug discovery Thema der Disputation:
Self-supervised learning for high-content imaging in drug discovery
Machine learning-driven data integration for drug discovery Thema der Disputation:
Self-supervised learning for high-content imaging in drug discovery
Abstract: “Self-supervised learning (SSL) has established itself as a training paradigm for learning meaningful representations from unlabeled data. SSL methods can extract features without costly and time-consuming manual annotations by using pretext tasks that leverage the intrinsic structure of data. Common SSL approaches include reconstruction-based methods like autoencoders, contrastive learning that groups similar samples while separating dissimilar ones, and self-distillation that learns consistent representations across augmentations of the data (Balestriero et al., 2023). SSL methods are particularly useful in high-content imaging (HCI), a domain in which large volumes of unlabeled fluorescence microscopy data are continuously generated. In this presentation, I will provide an overview of SSL approaches and illustrate HCI applications, including phenotype classification and mechanism of action identification, while also discussing the impact of these methods on drug discovery (Kraus et al., 2024; Kim et al., 2025).”
Time & Location
Feb 13, 2026 | 02:00 PM
Seminarraum 019
(Fachbereich Mathematik und Informatik, Arnimallee 3, 14195 Berlin)