Learning the Limits: How Data, Diversity, and Representation Control Machine-Learning Predictions of Reorganisation Energy

Key Points

• The aim is to determine how data diversity and representation influence the predictions of hole and electron reorganisation energies in organic semiconductors.
• Analyzed data sets of organic semiconductor properties
• Utilized machine learning algorithms to predict reorganisation energies
• Assessed the impact of data diversity on prediction accuracy
• Demonstrated that increased data diversity improves predictive accuracy
• Found significant variations in predictions based on representation of training data
• Observed that machine learning outperforms traditional calculation methods under certain conditions