Coding Exercise on Metrics of Interpretability
Learn to implement and calculate the sign agreement metric that measures how well post-hoc explanations align with ground-truth explanations by comparing the top-k features and their contribution directions. Understand the coding steps to quantify interpretability and assess explanation quality in deep learning models.
We'll cover the following...
We'll cover the following...
Problem statement: Sign agreement
Similar to the Top-k feature agreement metric, the sign agreement (SA) metric computes the fraction of Top-k features that are not only common between a given post-hoc explanation and the corresponding ground-truth explanation but also share the same sign (direction of contribution).
In other words, given a post-hoc explanation