A dropout uncertainty neural network (DUNN) model driver v001
Abstract
A dropout uncertainty neural network (DUNN) potential model driver, which supports running in both fully-connected mode and dropout mode. The DUNN can be used easily to quantify the uncertainty in atomistic simulations and determine the transferability of potential.
Fields of Study & Subjects
Machine Learning in Materials Science
Advanced Chemical Physics Studies
Model Reduction and Neural Networks
Publication details
- Journal / source
- Open Knowledgebase of Interatomic Models (OpenKIM)
- Publication date
- 1 January 2026