🩺 Edge#213: Testing Trained Models

the fundamental types of tests that can be applied to trained models +how Meta uses Bayesian Optimization for A/B tests +TensorFlow’s What-If Tool

In this issue:

• we overview the fundamental types of tests to be applied to trained models;

• we explain how Meta uses Bayesian Optimization to conduct better experiments in ML models;

• we explore TensorFlow’s What-If Tool, one of the most commonly used testing tools in the machine learning space.   

Enjoy the learning!  

---

💡 ML Concept of the Day: Testing Trained Models 

In our introduction to machine learning (ML) testing (Edge#209), we reviewed two fundamental approaches: pre-training and post-training testing. Today, we would like to dive deep into the methods for testing trained models. Most testing techniques for trained models fit into some of the following categories: