Unraveling learning characteristics of transformer models for molecular design
The learning characteristics of transformer based models for generative compound designs are studied using control calculations and careful manipulation of datasets.
Cheminformatics
Protocol to calculate and compare exact Shapley values for different kernels in support vector machine models using binary features
A protocol to calculate and compare exact Shapley values for support vector machine models with commonly used kernels and binary input features is developed.
Machine learning models with distinct Shapley value explanations decouple feature attribution and interpretation for chemical compound predictions
A method for the exact calculation of Shapley Values for Support Vector Machines is introduced and tested for compound prediction tasks.
Relationship between prediction accuracy and uncertainty in compound potency prediction using deep neural networks and control models
The prediction accuracy and uncertainty qunatification of deep neural networks and other control methods is compared using compound potency prediction tasks.