| About this Abstract |
| Meeting |
MS&T24: Materials Science & Technology
|
| Symposium
|
Materials Informatics for Images and Multi-dimensional Datasets
|
| Presentation Title |
Autonomous Approaches for Determining Structure-Processing-Property Relationships in Materials |
| Author(s) |
Rama Krishnan Vasudevan, Sumner Harris, Yongtao Liu, Arpan Biswas |
| On-Site Speaker (Planned) |
Rama Krishnan Vasudevan |
| Abstract Scope |
The use of traditional statistical and machine learning approaches have shown to be highly promising for understanding and predicting structure-processing-property relationships in materials, with methods including matrix factorization, random forests, and deep neural networks showing strong effectiveness at a range of tasks. However, the ability to rapidly learn these properties and change the way experiments are performed on the basis of these correlations is less explored.
Here, we will review our autonomous efforts in both scanning probe microscopy (SPM) as well as pulsed laser deposition (PLD), with Bayesian methodologies used to quickly optimize targets whilst simultaneously providing a quantifiable prediction of the structure-property relationships. More flexible kernel methods and their extensions to both SPM and PLD will be discussed, which can be used to reduce the number of overall experiments compared to more inflexible kernels. Extensions to incorporate robustness of the solution into the Bayesian framework will be discussed. |