Sign in to confirm you’re not a bot
This helps protect our community. Learn more

How to detect silent failures in ML models

Microsoft Reactor
115K subscribers
<__slot-el>
<__slot-el>
204 views Streamed 2 years ago
What is the session about? The session will teach you how to detect silent ML model failure without accessing the target data. We will cover the most likely causes for ML failure, like data and concept drift. You'll learn the tools (both statistical and algorithmic) used in detecting and dealing with these failures, their applications, and their limits. Who is it aimed at? This session is aimed at people with real-life experience with Data Science. Why should you attend? By the end of the session, you will be able to:  …...more
...more
How to detect silent failures in ML models
What is the session about? The session will teach you how to detect silent ML model failure without accessing the target data. We will cover the most likely causes for ML failure, like data and concept drift. You'll learn the tools (both statistical and algorithmic) used in detecting and dealing with these failures, their applications, and their limits. Who is it aimed at? This session is aimed at people with real-life experience with Data Science. Why should you attend? By the end of the session, you will be able to:
  • Detect drop in performance of their ML models without using ground truth data
  • Monitor their Machine Learning models
  • Understand and detect data drift to resolve the issues found
Any prerequisites? Familiarity with basic data science Bio Wojtek Kuberski is a co-founder of NannyML, an OSS Python library for detecting silent model failure. He holds a Master's Degree in AI. He previously founded and grew an ML consultancy. He likes tennis, chess, and most of all, food. [eventID:16559]

Follow along using the transcript.

Microsoft Reactor

115K subscribers
Live chat replay is not available for this video.