Benjamin Sanchez-Lengeling from Harvard University and Alán Aspuru-Guzik from the University of Toronto have successfully applied machine learning models to speed up the materials discovery process. Their paper Inverse molecular design using machine learning: Generative models for matter engineering was published July 27 in Science Vol. 361.
The authors propose a novel approach to both simplify the current paradigm and accelerate development: Generative models abstract and capture representations from original molecular data, while models such as variational auto-encoders, reinforcement learning and adversarial training are applied to guide the optimization process towards desirable qualities.
The traditional materials exploration process was greatly simplified after machine learning methods were applied. For example, battery material development usually required four steps with associated feedback before a product prototype can be produced: material concept, molecular synthesis, device construction, and testing. These steps can waste time on back-and-forth communication and experiment adjustments.
With the introduction of machine learning and inversion design, the new “closing the loop” paradigm can collect feedback immediately and optimize the exploration process.
Although the authors believe the representation of molecular structures based on generative models and encoder methods can achieve better inversion results, the research remains at an early stage.
The paper can be accessed here: http://science.sciencemag.org/content/361/6400/360
Author: Alex Chen | Editor: Michael Sarazen
0 comments on “Harvard & University of Toronto Researchers Apply Deep Generative Models to Inverse Molecular Design”