Model Distillation for Faithful Explanations of Medical Code Predictions

Model Distillation for Faithful Explanations of Medical Code Predictions

Explore knowledge distillation techniques for generating faithful and plausible explanations in machine learning models, particularly in clinical medicine and high-risk settings. Delve into Isabel Cachola's research from Johns Hopkins University's Center for Language & Speech Processing, focusing on improving interpretability of models with excellent predictive performance. Learn how this approach can support integrated human-machine decision-making and increase trust in model predictions among domain experts. Examine the application of these techniques to medical code predictions, based on the paper presented at the BioNLP 2022 workshop.