Accelerating Materials Discovery with AI and ML скачать в хорошем качестве

Accelerating Materials Discovery with AI and ML

Materials form the backbone of our society, presenting potential to at least 10/17 sustainable development goals. Traditional materials discovery relies on trial and error approaches thereby leading to a design to deploy period of 20-30 years. To address this challenge, in this talk, we will discuss the application of artificial intelligence (AI) and machine learning (ML) in accelerating materials modeling and discovery. Specifically, three aspects where AI and ML can be used include: (i) natural language processing (NLP) for extracting information from the materials literature, (ii) data-driven materials modeling, (iii) physics-informed machine learning for accelerated materials modeling. To demonstrate these aspects, three problems will be discussed. First focuses on extracting information on glasses and other materials from literature to answer specific queries. We will also discuss on how MatSciBERT, the first materials-aware language model, can be used to extract information regarding composition-property from the glass literature. Second, we focus on developing interpretable ML models for predicting properties based on the information extracted from the literature. This work covers nearly the entire periodic table for glass forming elements. Third, we will discuss on how to accelerate simulations using physics-informed ML (PIML). Here, we will discuss how interaction laws in nature can be discovered directly from the trajectory of physical systems using PIML. Altogether, the talk will cover various aspects of AI and ML that have been used to accelerate materials discovery. Finally, a brief outlook on the future prospects will be discussed. Anoop completed his Ph.D. in Civil Engineering from Indian Institute of Science Bangalore in 2015, after which, he worked as a postdoctoral researcher in University of California Los Angeles from 2015 to 2017. Shadad Hussain is the moderator.