BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Hybrid Multi-Modal Fusion for Heterogeneous Biomedical Data - Kons
 tantin Hemker (University of Cambridge)
DTSTART:20231121T130000Z
DTEND:20231121T140000Z
UID:TALK206245@talks.cam.ac.uk
CONTACT:Mateja Jamnik
DESCRIPTION:Technological advances in medical data collection such as high
 -resolution histopathology and high-throughput genomic sequencing have con
 tributed to the rising requirement for multi-modal biomedical modelling\, 
 specifically for image\, tabular\, and graph data. Most multi-modal deep l
 earning approaches use modality-specific architectures that are trained se
 parately and cannot capture the crucial cross-modal information that motiv
 ates the integration of different data sources. This talk presents the Hyb
 rid Early-fusion Attention Learning Network (HEALNet) – a flexible multi
 -modal fusion architecture\, which a) preserves modality-specific structur
 al information\, b) captures the cross-modal interactions and structural i
 nformation in a shared latent space\, c) can effectively handle missing mo
 dalities during training and inference\, and d) enables intuitive model in
 spection by learning on the raw data input instead of opaque embeddings. W
 e conduct multi-modal survival analysis on Whole Slide Images and Multi-om
 ic data on four cancer cohorts of The Cancer Genome Atlas (TCGA). HEALNet 
 achieves state-of-the-art performance\, substantially improving over both 
 uni-modal and recent multi-modal baselines\, whilst being robust in scenar
 ios with missing modalities.\n\n"You can also join us on Zoom":https://cam
 -ac-uk.zoom.us/j/92041617729\n
LOCATION:Lecture Theatre 2\, Computer Laboratory\, William Gates Building
END:VEVENT
END:VCALENDAR