Speakers
Description
This presentation will introduce the Hybrid ML research area, which aims to integrate deep learning with structured knowledge from mathematics and the natural and social sciences.
Hybrid ML is guided by the observation that both mathematics and the sciences can be seen as generators of compressed pattern representations. We will explain that the central goal of Hybrid ML is to align the representations inferred by deep learning models with those derived from theoretical frameworks. This goal is motivated by a fundamental hypothesis: that such aligned representations are not only easier to learn, but also more likely to contain optimal solutions (i.e., robust, interpretable, and efficient representations) tailored to the target objective. We will illustrate how this alignment can be implemented in practice, drawing on recent breakthrough results from the Hybrid ML area, and conclude with an overview of the collaboration channels available to Lamarr researchers.
Following this introduction, Paul Roetzer will give a spotlight presentation on their recent ICCV paper, “Fast Globally Optimal and Geometrically Consistent 3D Shape Matching”, in which they align neural representations with novel surface representations - in terms of cyclic paths - to obtain global, geometrically consistent matchings between three-dimensional shapes.
| Area Presenter | Ramsés Sánchez |
|---|---|
| Spotlight Presenter | Paul Roetzer |
