Yield Prediction of Organic Reactions in Biased Datasets via Positive-Unlabeled Learning

Jan Christopher Spies, Florian Boser, Frank Glorius

Submission Type: I want my submission to be considered for both oral and poster presentation.

Keywords: Molecular Machine Learning, Organic Chemistry, Yield Prediciton, Data Augmentation, Imbalanced Datasets

TL;DR: ML models trained on literature data are often limited by a lack of negative examples. We show that PU learning can be used to find "reliable negatives" from unlabeled data, enabling the training of robust yield prediction models on biased datasets.

Confirmation Of Submission Requirements: I submit an abstract. It uses the template provided on the submission page and is no longer than 2 pages.

Submission Number: 264