Fig. 1: Confusion matrix and confidence analysis for crystal system classification using the PE-AG-GMoE model.
From: Accelerating electron diffraction analysis using graph neural networks and attention mechanisms
a A confusion matrix from individual diffraction pattern predictions reveals the distribution of class assignments across symmetry groups, b aggregated predictions across multiple zone axes show a more consolidated assignment pattern with reduced inter-class confusion, c confidence scores for individual predictions reflect varying levels of certainty across different symmetry types, and d aggregated confidence scores demonstrate enhanced prediction consistency after multi-view integration.
