Trade-off Between Optimality and Explainability in Machine Learning Models

Explore the critical trade-off between model explainability and accuracy in machine learning through this insightful 44-minute talk by Nima Safaei, Senior Data Scientist at Scotiabank. Delve into the challenges of using black box models in high-risk areas due to lack of explainability, and examine the two-fold nature of explainability in ML: Causal Explainability and Counterfactual Explainability. Gain a deeper understanding of Counterfactual Explainability from an optimization perspective, and learn how post-optimality analysis can be applied to machine learning models. Investigate the limitations of optimization algorithms in guaranteeing global optimum solutions and how this impacts model explainability. Engage in a critical discussion on the trade-off between explainability and accuracy during model selection, considering whether a more explainable but less accurate model is preferable to a less explainable but more accurate one.

Trade off between Optimality and Explainability