Additive Feature Attribution

Additive feature attribution (AFA) methods aim to explain the predictions of machine learning models by assigning importance scores to individual input features, providing insights into how these features contribute to the model's output. Current research focuses on improving the accuracy and efficiency of AFA algorithms, such as SHAP and LIME variants, particularly for complex models like deep neural networks and ensembles, and addressing challenges like uncertainty quantification and handling feature interactions. This work is crucial for building trust in machine learning models across diverse fields, from fluid dynamics and heat transfer to online assessment and federated learning, by enhancing model interpretability and facilitating better understanding of model behavior.