Machine Learning for Materials Design
(Cangi Lab)
  • News
  • Research
  • People
  • Publications
  • Teaching
  • Contact
  • Internal

ICML 2024

conference
Karan Shah presents recent work at the ICML Workshop AI for Science: Scaling in AI for Scientific Discovery
Published

July 24, 2024

Image for Karan Shah presents recent work at the ICML Workshop AI for Science: Scaling in AI for Scientific Discovery

Karan’s recent work on accelerating electron dynamics simulations using neural networks was accepted as a poster at the ICML workshop “AI for Science: Scaling in AI for Scientific Discovery”, which took place on July 26 in Vienna, Austria.

Time-dependent density functional theory (TDDFT) is a common method used to study how electrons behave when exposed to laser fields. In our recent work, we introduced a new way to speed up these simulations using a special type of machine learning model. This model, called an autoregressive neural operator, helps predict the behavior of electrons over time. By incorporating physics-based rules and using detailed training data, our model is more accurate and faster than traditional methods. We tested our approach as a proof of concept on simple toy models and showed that it works well. This method could make it easier to model how molecules and materials respond to lasers in real-time experiments.

Read more about this in our preprint.

 

© 2025 Attila Cangi · Impressum · Built with Quarto