Virtual Talk: Leveraging small datasets for molecular machine learning

Ron Dror
Stanford University

The most striking successes of machine learning have been in domains where large amounts of data are available for training. In many important problems in structural biology and drug discovery, however, only small amounts of data are available. Indeed, computational prediction is particularly valuable for types of data that are difficult to collect experimentally. I will discuss several strategies for effective machine learning given small datasets of molecular structures or properties.

Back to Learning and Emergence in Molecular Systems