Machine learning is a class of methods that rely on existing data to train a model and then predict a quantity of interest. Some machine learning methods can also estimate the uncertainty in those predictions. My research program is pioneering the exploitation of this uncertainty as an input for the second machine learning model to predict a different and final physical property.
This generic technology should find an application across a broad range of physical and biological systems. For example, motivated by Renormalization Group Theory, in which fluctuations determine the macroscopic state of the system, the methodology could predict heat capacity of new materials. A second example could be the replication, growth, and death of cells in the human body.
In Plain English
We use computers to understand patterns in data. These patterns are often hard to notice because of the noise. However, we do not get rid of the noise to see the patterns. Instead, we use the noise to see these patterns in greater detail. This can help us understand different phenomena in Physics and Biology.
Magnetotransport in semiconductors and two-dimensional materials from first principles
Phys. Rev. B 103, L161103 (Letter, Editors' suggestion)