Facebook has launched a 6-part video series dedicated to providing practical tips about how to apply machine-learning capabilities to real-world problems. The Facebook Field Guide to Machine Learning is developed by the Facebook ads machine learning team.
The development process of an ML model
Source: Facebook research
The video series will cover how the entire development process works. This includes what happens during the training of machine learning models and what happens before and after the training process in each step. Each video also includes examples and stories of non-obvious things that can be important in an applied setting. The video series breaks down the machine learning process into six steps:
It is necessary to have the right set up before you go about choosing an algorithm. The first video, Problem definition, talks about how to best define your machine learning problem before going into the actual process. You save almost a week’s worth of time by spending just a few hours at the definition stage.
This tutorial teaches developers how to prepare the training data. The training data is a powerful variable to create high-quality machine learning systems.
The third lesson talks about the steps to evaluate the performance of your machine learning model.
The fourth tutorial explains examples of various features such as categorical, continuous and derived features. It also describes how to choose the right feature for the right model. The video also talks about changing features, feature breakage, leakage, and coverage.
The next lesson describes how to choose the right machine learning model for your data and find the algorithm to implement and train that model. It also offers tips and tricks for picking, tuning and comparing models.
The final tutorial covers experimentation, which is about making your experiments actionable. A large part of the tutorial is dedicated to the difference between offline and online experimentation.
The entire video series is available on the Facebook blog for you to watch.