To use machine learning to tackle the greatest challenges in the physical sciences, we need methods built to handle the “data types” of physical systems: geometry and the geometric tensors. These are traditionally challenging data types to use for machine learning because coordinates and coordinate systems are sensitive to the symmetries of 3D space: 3D rotations, translations, and inversion.
In this talk, I present a method that I have been developing with my colleagues for the past three years, Euclidean neural networks. These networks preserve Euclidean symmetry by construction, making them incapable of unphysical bias due to a change of coordinates. They eliminate the need for data augmentation -- the 500-fold increase in brute-force training necessary for a model to learn 3D patterns in arbitrary orientations. This makes them extremely data-efficient; they result in more accurate models and require less training data to do so, which is ideal for modeling from scientific data that is expensive, difficult to acquire, or highly-varied.
Most symmetry-aware machine learning models in the physical sciences avoid augmentation through overly restrictive invariance, only using scalar information and throwing away coordinate systems altogether. But this comes at a price; many of the rich consequences of Euclidean symmetry are lost: geometric tensors, point groups, space groups, degeneracy, atomic orbitals, etc. Many of these consequences are directly responsible for the behavior of physical systems. In contrast, Euclidean neural networks treat the full equivariance of physical systems and thus can more faithfully describe physical systems.
I describe how Euclidean neural networks work, demonstrate their effectiveness on a variety of real-world tasks, and introduce new capabilities my colleagues and I are developing with these methods. I also show how to efficiently and flexibly build equivariant models using our open-source PyTorch package e3nn (https://e3nn.org).
Dr. Tess Smidt is the 2018 Alvarez Fellow in Computing Sciences at Lawrence Berkeley National Laboratory where she designs neural networks from first principles for rich data types such as geometry and scientific data. She earned her PhD in Physics from UC Berkeley in 2018 with Prof. Jeffrey B. Neaton using density functional theory to understand structure-property relations underlying quantum magnetism, ferroelectricity, and quantum confinement. In the final year of her PhD, she interned with the Google Accelerated Science team and developed neural networks that have Euclidean symmetry built into them, making them extremely efficient at learning from 3D data. As an undergraduate at MIT majoring in physics and minoring in architecture, Tess engineered giant neutrino detectors in Prof. Janet Conrad’s group and created a permanent science-art installation on MIT's campus called the Cosmic Ray Chandeliers, which illuminate upon detecting cosmic-ray muons.
Hosts: Tayo Akinwande & Marc Baldo