Boosting AI Reliability- Uncertainty Quantification with MAPIE
summary
Library MAPIE implements conformal prediction methods that produce prediction intervals and sets with provable marginal coverage under exchangeability, model‑agnostic and integrable with scikit‑learn TensorFlow or PyTorch. It uses a held‑out calibration split (suggested ~20%) or cross‑/jackknife+ variants, offers conformalized quantile regression and subgroup guarantees (MREA), and demonstrates practical numbers — e.g. an 80% target yields ~60% singleton predictions and ~80% sets of size <3 — with notebooks and code on GitHub. MAPIE split conformal and jackknife+ 80% confidence → 60% singletons
AI-written summary · about
chapters (6)
- 0:00Uncertainty quantification intro
- 3:14Why conformal predictions
- 9:20MAPIE implementation
- 12:54MAPIE use case
- 19:23MAPIE roadmap
- 22:43Q&A
related talks
49mBoosting AI Reliability: Uncertainty Quantification with MAPIE
01h 24mBoosting AI Reliability: Uncertainty Quantification with MAPIE
44mAccuracy Is Not Enough: Building Trustworthy AI with Conformal Prediction
30mConformal Prediction: uncertainty quantification to humanise models
31mUncertainty quantification: How much can you trust your machine learning model?
35mUncertainty Unleashed
34mPrediction intervals
19mUnderstanding, Generating, and Evaluating Prediction Intervals