Reliable Machine Learning

Description

In this course, you will :

• An awareness of the many types of testing and their usefulness in machine learning applications.
• Understanding of how to use Pytest to improve the resilience of machine learning systems.
• An in-depth overview of the best (and worst) testing practises.
• Hands-on experience troubleshooting machine learning applications.

Syllabus :

1. Introduction to Reliable ML

• Who Should Take This Course?
• Why Do We Need Tests?
• Various Software Testing Paradigms

2. Software Testing

• Unit Testing
• Introduction to pytest
• Fixtures in pytest
• Integration Testing
• Mocks and Patches
• Mutation and Property-Based Tests

3. Best and Worst Practices

• Test-Driven Development
• Flaky and Negative Tests
• Automation of Testing

4. ML-Specific Tests

• Why Machine Learning is Specific
• Behavioral Testing
• Comparing the Environments
• Tests for Exact Values
• Smoke Tests for the ML Pipeline
• Fixtures and Mocks in Machine Learning Tests
• Samples as Tests

5. ML Software Reliability outside of Tests

• Runtime Checks
• Type Hinting
• Logging and Debugging
• ML Monitoring Guide

Show more