Conformal Prediction
Conformal prediction is a framework for quantifying the uncertainty of individual predictions made by any machine learning model. Instead of just outputting a single prediction, it provides a prediction region (for classification) or an interval (for regression) that is guaranteed to contain the true outcome with a user-specified probability. This guarantee holds true without making strong distributional assumptions about the data.
Rather than saying 'the temperature will be 25 degrees,' conformal prediction says 'the temperature will be between 23 and 27 degrees with 90% certainty.'
It allows AI systems to formally communicate their confidence and potential error margins for each prediction, crucial for high-stakes applications like medical diagnosis or autonomous driving.
Learn one new AI thing every day.
Daily Deck sends you seven plain-English cards like this every morning. Free.
Start free