Explore the challenges and solutions for implementing responsible AI in financial services through this 40-minute conference talk from the Toronto Machine Learning Series. Learn how to minimize risk and accelerate MLOps by leveraging machine learning monitoring and explainability techniques. Discover strategies for ensuring model performance, fairness, and compliance with regulations like SR 11-7 and OCC 2011-12. Gain insights into monitoring ML models at scale, documenting model behavior, and debugging complex models for quick issue resolution. Understand the importance of Explainable AI in illuminating black box models and addressing performance issues related to labels, drift, and data errors. Examine real-world applications of Model Performance Management (MPM) across the ML lifecycle, including a case study from a top 5 bank.
Minimize Risk and Accelerate MLOps with ML Monitoring and Explainability
Toronto Machine Learning Series (TMLS) via YouTube
Overview
Syllabus
Intro
Key Use Cases of ML In Finance
Models fail frequently
Most models are a black box
Regulations and Guidelines
MPM illuminates the black box
Catch Performance Issue with Labels
Catch Performance Issue with Drift
Catch Performance Issue with Data Errors
Catch Bias Issues
Solution - Explainability
Explaining a Prediction
Explanations - The Fed Remarks
Explaining a Segment or Model
Model Summary Report Powered by Explainability
Putting it together - Monitoring & Explainability
MPM Across the ML Lifecycle
Fiddler in Action: Top 5 Bank
Taught by
Toronto Machine Learning Series (TMLS)
Related Courses
-
Monitoring Machine Learning Concepts
-
Build High Performance MLOps With ML Monitoring and AI Explainability
-
Build High Performance MLOps With ML Monitoring and AI Explainability
-
Build High Performance MLOps With ML Monitoring and AI Explainability
-
Obtaining New Insights on Model Behavior with Fiddler - MLOps Meetup #110
-
ML Observability - A Critical Piece for Making Models Work in the Real World
