Explore the concept of passive symmetries in machine learning through this insightful one-hour lecture by Soledad Villar, Assistant Professor at Johns Hopkins University. Delve into the implications of arbitrary investigator choices in data representation and their corresponding exact symmetries. Examine the role of passive symmetries in machine learning, including permutation symmetry in graph neural networks, and learn about best practices in implementation. Discover the conditions for implementing passive symmetries as group equivariances and understand their connections to causal modeling. Gain valuable insights on how passive symmetries can enhance out-of-sample generalization in learning problems. This talk, part of the Data-Driven Physical Simulations (DDPS) webinar series, offers a unique perspective on the intersection of mathematics, physics, and machine learning.
Overview
Syllabus
DDPS | The passive symmetries of machine learning by Soledad Villar
Taught by
Inside Livermore Lab
Related Courses
-
Exact and Approximate Symmetries in Machine Learning Models
-
Introduction to Equivariant Machine Learning - IPAM at UCLA
-
Equivariant Machine Learning Structured Like Classical Physics
-
Learning Geometry and 3D Symmetries - Day 1
-
Understanding Quantum Machine Learning Also Requires Rethinking Generalization
-
Subgraph-Based Networks for Expressive, Efficient, and Domain-Independent Graph Learning
