About this Abstract |
| Meeting |
2026 TMS Annual Meeting & Exhibition
|
| Symposium
|
Accelerated Discovery and Insertion of Next Generation Structural Materials
|
| Presentation Title |
Leveraging Domain Knowledge for Optimal Initialization in Materials Optimization Frameworks |
| Author(s) |
Trevor Hastings, James Paramore, Brady Butler, Raymundo Arróyave |
| On-Site Speaker (Planned) |
Trevor Hastings |
| Abstract Scope |
Machine learning-based optimization strategies have emerged as an effective means to accelerate the discovery of new materials by efficiently exploring complex and high-dimensional design spaces. However, the success of optimization frameworks greatly depends on how well the campaign is initialized---the selection of seed data points from which the optimization starts. In this study, we focus on improving these initial datasets by incorporating materials science expertise into the selection process. Using Bayesian Optimization as a prototypical framework with real-world design criteria, we demonstrate that incorporating domain knowledge leads to more diverse initial datasets. These enhanced starting points significantly improve the efficiency of subsequent optimization efforts. We also introduce clear metrics for assessing the quality and diversity of initial datasets, providing a straightforward way to compare different initialization strategies. This approach acts as a widely applicable enhancement to various materials discovery scenarios. |
| Proceedings Inclusion? |
Planned: |
| Keywords |
Modeling and Simulation, Computational Materials Science & Engineering, Machine Learning |