What is it about?

Machine learning has found its way into many areas of science. In materials science, machine learning models can be used to predict the properties of new materials. This can be done even based on limited data. But it can be tricky to teach models how different properties from different fields are related. It is also difficult to train models to learn these links themselves. The authors of this study created a new framework to tackle these challenges. They have called it the Hierarchical Correlation Learning for Multi-property Prediction (H-CLMP) framework. This framework brings many machine learning methods together. First, it allows for prediction using only a material’s composition. Second, it can use this data to predict multiple properties at once for new materials. Finally, it can be trained with data from different domains. This is possible thanks to a strategy called transfer learning. It helps H-CLMP to draw the connections between a material’s composition and properties.

Featured Image

Why is it important?

Exploring new material compositions is a key part of materials science. However, doing this through trial and error is a costly and time-consuming process. Machine learning frameworks like H-CLMP can use existing data to predict the composition of new materials. These predictions can help guide scientists in the right direction when looking for materials with unique properties. Overall, this study highlights a new way in which artificial intelligence can speed up the progress of science. KEY TAKEAWAY: The H-CLMP framework will help researchers predict the properties of novel materials. The framework can also bring together knowledge from different areas of materials science.

Read the Original

This page is a summary of: Materials representation and transfer learning for multi-property prediction, Applied Physics Reviews, June 2021, American Institute of Physics, DOI: 10.1063/5.0047066.
You can read the full text:

Read

Contributors

Be the first to contribute to this page