Discovery of Bimetallic Catalysts with DFT-assisted Machine Learning Models used
- Random Forest Regression
- Gaussian Process Regressor
Evaluation Criteria: Combination of
- RMSE
- MAE
- Leave One-out-Cross Validation
Advantages of the Methods Used
-
The novel aspect of the study is that it allowed for the interpretation of experimental observations and the computational discovery of new catalysts, even with limited experimental data. The study demonstrated the potential for rational design of improved catalysts by understanding atomic-scale factors that control catalytic activity and selectivity, which is crucial for future advancements in catalyst discovery.
-
The combination of machine learning and first-principles calculations could be applied to other catalytic systems, expanding the scope of this approach in catalyst discovery.