The design patterns in this e book seize finest practices and options to recurring issues in machine studying. Authors Valliappa Lakshmanan, Sara Robinson, and Michael Munn catalog the primary tried-and-proven strategies to assist engineers sort out issues that steadily crop up throughout the ML course of. These design patterns codify the expertise of lots of of consultants into recommendation you’ll be able to simply observe.
The authors, three Google Cloud engineers, describe 30 patterns for knowledge and drawback illustration, operationalization, repeatability, reproducibility, flexibility, explainability, and equity. Each sample features a description of the issue, a wide range of potential options, and suggestions for selecting essentially the most acceptable treatment on your scenario.
You’ll learn the way to:
- Identify and mitigate widespread challenges when coaching, evaluating, and deploying ML fashions
- Represent knowledge for various ML mannequin sorts, together with embeddings, characteristic crosses, and extra
- Choose the precise mannequin sort for particular issues
- Build a sturdy coaching loop that makes use of checkpoints, distribution technique, and hyperparameter tuning
- Deploy scalable ML techniques you could retrain and replace to replicate new knowledge
- Interpret mannequin predictions for stakeholders and be sure that fashions are treating customers pretty