Question 3/10 v3 lecture 8

We have a model that overfits. What actions, and in what order, should we consider to maximize its performance?


get more data - augment data - search for generalizable architecture - add regularization - reduce architecture complexity


Relevant part of lecture

supplementary material

It is important to thoroughly consider how we can make our model perform better (getting more data, augmenting data, searching for an architecture that generalizes better) before we jump to limiting its capabilities. It's easy to overlook these steps and it comes at a price of limiting the performance our model can achieve.