Hands-On Artificial Intelligence for Beginners

Testing deep learning algorithms

The AI community is severely behind on adopting appropriate testing procedures. Some of the most advanced AI companies rely on manual checks instead of automating tests on their algorithms. You've already done a form of a test throughout this book; cross-validating our models with training, testing, and validation sets helps to verify that everything is working as intended. In this section, we'll instead focus on unit tests, which seek to test software at the smallest computational level possible. In other words, we want to test the little parts of our algorithms, so we can ensure the larger platform is running smoothly. 

Testing our algorithms helps us keep track of non-breaking bugs, which have become ubiquitous ...