Summary

Model interpretability is a relatively new area, with most publications happening in the last few years, but it is a very active area of research in DL now that is of utmost importance to realize the promise of precision medicine. The ability to interpret model decisions or predictions has several business advantages and can ultimately lead to higher profits. Because of model interpretability, more and more companies are leaning toward using DL models in their decision-making processes. This is not restricted to low-risk sectors but also high-risk sectors such as medicine and genomics too. If they are not currently using model interpretability, they plan to incorporate it into their future strategy.

This chapter is an attempt to introduce you to model interpretability, why it is important, why business organizations care about it, and different methods for performing model interpretability, specifically for black-box models such as DNNs in the genomics field. The chapter...