Researchers develop machine learning method to predict material properties with limited data

Researchers at the Indian Institute of Science and University College London have created a new machine learning method to predict material properties using limited data. This approach enhances material discovery, particularly for semiconductors. The method employs transfer learning, where a model is pre-trained on a large dataset and then fine-tuned for specific tasks. This allows for better predictions of material properties compared to models trained from scratch. The team's Multi-property Pre-Training framework showed strong results, accurately predicting band gap values for 2D materials not included in the training data. This technology may also aid in battery development and support India's semiconductor manufacturing efforts.