Explore machine learning testing techniques in this EuroPython conference talk. Delve into the importance of writing high-quality code for machine learning algorithms through automated testing. Examine the unique challenges of testing scientific code, including handling unstable data and avoiding under/overfitting. Learn about specific testing tools like numpy.testing for numerical data. Analyze famous machine learning techniques from a testing perspective, gaining deeper insights into learning model functionality. Suitable for intermediate Python programmers, this practical, code-oriented talk requires no prior knowledge of testing or machine learning algorithms. Cover topics such as linear regression, classification, clustering, supervised learning, unit testing, model performance evaluation, cross-validation, and confusion matrices.
Overview
Syllabus
Introduction
Agenda
Linear Regression
Classification
Clustering
Supervisor Learning
Mush Learning
Complexity
Risk
Fault
Unit Test
Number Problems
Newbie Testing
MachFramer
Function
Output
Model performance
Cross validation
Train Test Plate
Confusion Matrix
Recommendation
Taught by
EuroPython Conference
