Deep Graph Kernel Learning for Material & Atomic Level Uncertainty Quantification in Adsorption Energy Prediction

Key Points

• This research aims to enhance the prediction of adsorption energy using advanced graph neural networks to quantify uncertainty.
• Utilized graph neural networks as surrogates for density functional theory calculations.
• Applied kernel learning techniques to assess uncertainty in predictions.
• Evaluated the model's accuracy in predicting adsorption energy.
• Successfully improved prediction accuracy for adsorption energy.
• Demonstrated effective uncertainty quantification in the predictions.
• Showed potential for accelerating the discovery of catalytic materials.