Researchers at the Indian Institute of Science (IISc), in collaboration with University College London, have developed machine learning-based methods to predict material properties with limited data.

According to the research, this method can help in semiconductor discovery. Additionally, it can forecast the speed at which ions can flow between battery electrodes, which aids in the development of more advanced energy storage technologies.

For the study, the research team created a model based on Graph Neural Networks (GNNs), under the direction of Sai Gautam Gopalakrishnan, an assistant professor in the Department of Materials Engineering.

According to the IISc, one obstacle is the dearth of material property data, which is required to train models that can identify the kinds of materials that have particular characteristics like electronic band gaps, formation energies, and mechanical properties. This is a result of the time-consuming and costly techniques now in use.

“In transfer learning, researchers use a large model first pre-trained on a large dataset and then fine-tuned to adapt to a smaller target dataset. In this method, the model first learns to do a simple task like classifying images into, say, cats and non-cats, and is then trained for a specific task, like classifying images of tissues into those containing tumors and those not containing tumors for cancer diagnosis,” says Sai Gautam.

“The architecture of the GNN, such as the number of layers and how they are connected, determines how well the model can learn and recognize complex features in the data,” says IISc scientists.

After pre-training and fine-tuning their transfer learning-based model, the IISc team discovered that it outperformed models that were built from scratch. The group also pre-trained their model on seven distinct bulk 3D material characteristics at the same time using a framework known as Multi-property Pre-Training (MPT).