scientific-uncertainty-quantification

What problem does it solve?

This skill quantifies predictive uncertainty in machine learning, providing calibrated probabilistic estimates and reliable prediction intervals to increase trust in model outputs.

Core Features & Use Cases

• Conformal prediction for distribution-free prediction intervals in regression and classification.
• MC dropout and deep ensemble methods to separate epistemic and aleatoric uncertainty.
• Calibration analysis and ECE estimation to assess probability calibration across models.
• ToolUniverse integration and benchmarking utilities for reproducible uncertainty quantification.

Quick Start

Run the quick start scenario to quantify uncertainty in a trained model and generate prediction intervals.