NLDL special session on geometry and topology aware deep learning

Papers are invited for a special session on geometry and topology aware deep learning to be organized as a part of the Northern Lights Deep Learning Conference - NLDL - which takes place in Tromsø, Norway, Jan 10-12, 2022. The special session focuses on leveraging the geometry and topology of different problems to improve deep learning approaches. The underlying idea behind geometry and topology aware deep learning is that learning complicated representations is easier and more effective if the architecture can leverage the geometric and topological properties of a problem. Using our knowledge in geometry and topology simplifies learning problems. Some examples are topological regularization, geometric interpretation, faster training, and task generalization. Expected contributions will cover deep learning methods based on ideas in geometry and topology, as well as their applications.

We are accepting two alternatives for contributions: (1) Full paper submissions (6 pages) will be presented and will be published in the conference proceedings**; (2) Extended abstracts (2 pages) will be presented (but not published in the conference proceedings). The review process is double-blind. Please submit your papers via https://septentrio.uit.no/index.php/nldl/about/submissions (for full papers) and https://cmt3.research.microsoft.com/NLDL2022/ (for extended abstracts) . For more information on NLDL, please visit http://nldl.org. Inquiries about the special session should be directed to me.

MLVis presentation

Natacha Galmiche presents our paper “Revealing Multimodality in Ensemble Weather Prediction” at EuroVis 2021.

CEDAS poster

Erlend Raa Vågset received a best poster award at the CEDAS conference for our poster on “Finding Geometrically Concise Representations of Homology” based on this preprint.