Explore a comprehensive 84-minute conference talk from the Toronto Machine Learning Series that delves into conformal prediction and uncertainty quantification in machine learning models. Learn from MLBoost President Mahdi Consent as he demonstrates how to create statistically sound uncertainty intervals for model predictions using distribution-free validity approaches. Master practical techniques for implementing conformal prediction across computer vision, natural language processing, and deep reinforcement learning applications. Gain hands-on experience through Python code samples and Jupyter notebooks while discovering methods to handle structured outputs, distribution shifts, and outliers. Understand how to leverage pre-trained neural networks to generate reliable uncertainty sets with customizable confidence levels, essential for high-stakes applications in medical diagnostics and industrial AI. Walk away equipped with robust tools and methodologies to ensure model reliability and trustworthiness in complex machine learning challenges.
Uncertainty Quantification with Conformal Prediction - A Path to Reliable ML Models
Toronto Machine Learning Series (TMLS) via YouTube
Overview
Syllabus
Uncertainty Quantification with Conformal Prediction: A Path to Reliable ML Models
Taught by
Toronto Machine Learning Series (TMLS)
Related Courses
-
Uncertainty Quantification in Machine Learning: Federated Conformal Prediction
-
Quantifying the Uncertainty in Model Predictions Using Conformal Prediction
-
Uncertainty Quantification with Physics-Informed Machine Learning
-
The Materials Simulation Toolkit for Machine Learning (MAST-ML): Automating Development and Evaluation of Machine Learning Models for Materials Property Prediction
-
Uncertainty Quantification and Deep Learning for Water-Hazard Prediction
