We also introduce a method for rotating latent factors to enhance interpretability with respect to binary features. FACTM was validated on text and video benchmarks as well as real-world music and COVID-19 datasets.

FACTM is a multi-view and multi-structure Bayesian model that combines factor analysis with correlated topic modeling and is optimized using variational inference. The model is integrating simple vector views, as in traditional FA, with structured views, which have multiple data points per sample.

Keynotes:
@machine.learning.bio Christian Dallago, Duke University, US
@delafuentelab.bsky.social  Cesar de la Fuente, UPenn, US
@fragrisoni.bsky.social  Francesca Grisoni - TU/e, NL
Birte Hoecker, Bayeruth University, DE

The Symposium will explore recent advancements in generative AI models for molecule discovery, with a focus on applications in health and drug development. Topics include AI-driven design of small molecules, peptides, and proteins, drug repurposing, and AI-based modelling of DNA and RNA sequences.

2. Explore multimodal, high-dimensional, spatial data of tumors, with collaboration partners in the European Mission Cancer project SPACETIME, developing AI models for Lung cancer.
* Advertisement: euraxess.ec.europa.eu/jobs/329762
* More about SPACETIME: cordis.europa.eu/project/id/1...

1. Work on our @erc.europa.eu funded project DOG-AMP and design novel Generative AI methods for antimicrobial peptide discovery!
* Advertisement: euraxess.ec.europa.eu/jobs/329758
* More about the project: www.youtube.com/watch?v=_Y-_...

His project is a follow up to his recent publication "CaClust: linking genotype to transcriptional heterogeneity of follicular lymphoma using BCR and exomic variants" in Genome Biology rdcu.be/d2x0b