Monday, April 22, 2024
1:30 pm – 5:00 pm
Summit - Seattle Convention
Center, Level 3, Room 344
Instructor: Ming Hu, University of South Carolina
The advent of machine learning
(ML) and artificial intelligence (AI) has revolutionized many aspects of modern
science and technology and has sparked significant interest in the materials
science community in recent years. Despite some early deployment of ML/AI in
the thermal science area, the power of AI has not been maximized. Existing ML
methods for predicting phonon properties of crystals are limited to either a
small amount of training data or a material-to-material basis, primarily due to
the exponential scaling of model parameters with the number of atomic species
or elements. This renders high-throughput infeasible when facing large-scale
new materials.
This tutorial lecture will
introduce some state-of-the-art ML/AI approaches for predicting energy
carriers' transport behaviors in materials. Both traditional ML methods (such
as random forest) and novel graph neural networks will be presented with showcase
studies on training, testing and prediction of lattice vibrations (phonons).
Particular focus will be our recently developed Elemental Spatial Density
Neural Network Force Field (Elemental-SDNNFF) with abundant atomic-level
environments as training data. Benefiting from the innovative architecture of
the algorithm, sub-trillion atomic data can be integrated to train a single
deep neural network for predicting complete phonon properties of >100,000
inorganic crystals spanning 63 elements in the periodic table.
We will also illustrate recent
ML/AI algorithms for discovering promising thermal materials for various energy
applications, including but not limited to thermoelectrics, interfacial thermal
management, topological phonons for quantum information technology.
Tutorial Schedule
1:30 pm
Machine Learning Toward
Advanced Thermal Materials—Algorithm
Ming Hu, University of South
Carolina
2:45 pm BREAK
3:15 pm
Machine Learning Toward
Advanced Thermal Materials—Applications
Ming Hu, University of South
Carolina