Zico Kolter - Incorporating physics and decision making into deep learning via implicit layers

Talk starts at 2:57

Prof. Zico Kolter from Carnegie Mellon speaking in the Data-driven methods for science and engineering seminar on January 22, 2021.

For more information including past and upcoming talks, visit: http://www.databookuw.com/seminars/

Sign up for notifications of future talks: https://mailman11.u.washington.edu/mailman/listinfo/datadriven-seminar

Links:
https://zicokolter.com/
http://implicit-layers-tutorial.org/

Abstract: Despite their wide applicability, deep networks often fail to exactly capture simple "known" features of real-world data sets, such as those governed by physical laws. In this talk, I'll present methods for integrating hard constraints, such as those associated with decision making, optimization problems, or physical simulation, directly into the predictions of a deep network. Our tool for achieving this will be the use of so-called implicit layers in deep models: layers that are defined implicitly in terms of conditions we would like them to satisfy, rather than via an explicit computation graph. I'll discuss how we can use these layers to embed (exact) physical constraints, robust control criteria, and task-based objectives, all within the framework of traditional deep learning. I will highlight several applications of this work in reinforcement learning, control, energy systems, and other settings, and discuss generalizations and directions for future work in the area.

Видео Zico Kolter - Incorporating physics and decision making into deep learning via implicit layers канала Physics Informed Machine Learning