ML+X Forum: Representations for Learning

How can researchers exploit the structure of their data to train more accurate and reliable deep learning models (e.g., via geometric deep learning which generalizes neural networks to Euclidean and non-Euclidean domains, such as graphs, manifolds, meshes, or string representations)? Similarly, how does the performance and design of different artificial neural network models (e.g., CNNs, vision transformers, CLIP) relate to a model’s learned feature representations? Do some models come closer to mirroring human visual representations than others, and can this observed relationship be exploited to develop better performing models?

These questions and more will be explored at April 11th’s ML+X forum, 12-1pm CT! Presenters include Daniel McNeela (Biomedical Data Science MS and incoming CS PhD student) and Kushin Mukherjee (Psychology PhD student)

1. Geometric Deep Learning for Molecules and Biological Networks — Daniel McNeela

2. Characterizing the Building Blocks of Human Perceptual Similarity in Abstract Drawings Using Deep Neural Networks — Kushin Mukherjee

RSVP for the post-event social: Want to discuss ML projects and connect with the presenters following this event? Drop in at the ML Community's monthly social — ML+Coffee (April 12th, 9-11am). Learn more and register.

Finding the Orchard View room: The Orchard View room is located on the 3rd floor of Discovery Building — room 3280. To get to the third floor, take the elevator located next to the Aldo’s Cafe kitchen (see photo). If you cannot attend in-person, we invite you to stream the event via Zoom.

Join the ML Community google group: The ML Community has a google group it uses to send reminders about its upcoming events. If you aren't already a member of the google group, you can use this link to join. Note that you have to be signed into a google account to join the group. If you have any trouble joining, please email faciltator@datascience.wisc.edu.