I'm going to give a practice GAC meeting on Wednesday, Nov. 29 at 11am in the Division Room. Title and abstract are below.

If you'd like to come and help me practice, or are interested in hearing about the current state of molecular autoencoders, feel free to come.

Thanks!

Jennifer

Title: Semi-supervised learning for Molecular Autoencoders

Abstract:

Molecular autoencoders are capable of encoding molecules into a continuous vector representation. This encoding can then be used to map to properties, and even to optimize for properties. However, this autoencoder requires a large dataset in order to learn an accurate encoding. I will describe how semi-supervised learning can be used to help the molecular autoencoder learn from a smaller dataset.