The state-of-the-art models in XAI
XAI is a research area that has been gaining popularity in the past few years. In this section, we will look at a synthesis of the most critical XAI models in use today.
Accumulated Local Effects
The Accumulated Local Effects (ALE) method computes the effects of features globally. It is mainly used with tabular data, where different variables can be compared. The idea behind ALE is that if we have a small enough window, we can create an accurate estimate of the changes within a specific period. So, if we have a variable and we can sample its values across different periods, we can create an accurate estimate of how that variable is changing over time. The process is then repeated across all the accumulated data and is used to augment the global prediction. The algorithm focuses on the changes between one sampling point and the other, thus making the data relatively easy to interpret for any analyst.
Anchors
Anchors try to explain the behavior...
