In this talk, I will provide an update of my group's progress in generative models for molecules. In particular, I will describe a new molecular string representation with a 100% success rate in the generation of valid molecules called SELFIES. I will describe that armed with SELFIES, one can carry out interesting experiments of alternatives to GAN and autoencoders that provide significant speedups and provide other freedoms. As the work is unpublished, I will remain mysterious in the public abstract.
Back to Workshop I: From Passive to Active: Generative and Reinforcement Learning with Physics