ML application lifecycle
summary
A practical overview of the machine-learning application lifecycle covering concrete stages: problem formulation, data definitions, modeling, production system design and implementation, testing/deployment/maintenance, and online evaluation and evolution. The talk uses lessons learned and recommendations for each stage to show how to build and operate ML systems effectively.
AI-written summary · about
chapters (18)
- 0:00Intro
- 2:26Formal approach to engineering ML Systems
- 8:32System Objectives
- 10:57ML & Data Science Learning Programs
- 11:23Factors for Success of ML Systems
- 12:11Problem Formulation
- 14:12Reseller Fraud Example
- 17:34Key elements of an ML Prediction Problem
- 18:19Instance Definition
- 19:00Target Variable to be Predicted
- 21:38Input features
- 24:45Sources of Data
- 27:44Modeling Metrics - Classification
- 30:35Ethical and Fairness Constraints
- 31:52Deployment Constraints
- 33:46Data Definitions
- 35:48Offline Modeling
- 39:44Main Takeaways
related talks
52mThe ML Lifecycle in Combat Conditions
27mA Look Inside Applied Machine Learning
37mApplying ML in AdTech and Lifecycle of an ML project
30mChallenges of machine learning: The new ML lifecycle
30mThe State of Production Machine Learning in 2023
55mOrganizing the Machine Learning Lifecycle with MLflow
14mGetting Machine Learning applications into production using MLOps and CD4ML
47mMLOps tools to automate ML deployment