Applied AI Under Uncertainty
Calibration, causality and decisions in healthcare, finance, law and recommendation
-
- $8.99
-
- $8.99
Publisher Description
Every machine-learning model is wrong some of the time. This book teaches you to say how often, with a guarantee you can prove and code you can run. Applied AI Under Uncertainty, Volume 1, starts from the decision a prediction feeds and builds the toolkit to make that decision defensible: calibrated probabilities, prediction sets with finite-sample coverage guarantees, causal and off-policy evaluation, grounded retrieval for language models, and statistical evaluation under distribution shift.
The toolkit is then applied to four industries. In healthcare, you build a survival model whose risk estimates are calibrated, carry guaranteed intervals, and are audited for subgroup fairness. In finance, you learn why most backtests overstate performance and how to correct for it. In law and compliance, you build question answering with checkable citations and a bounded rate of unsupported answers. In recommendation and search, you evaluate a ranker offline when the logged data comes from an older policy.
Every chapter follows the same plan: the problem and its stakes, a precise formalization, the method with its derivation and proofs, a complete implementation in Python, a worked example with printed output, an evaluation with confidence intervals, the ways it fails in production, and exercises. All examples use synthetic data or open datasets, so every listing can be run and changed.
Written for engineers, data scientists, quantitative researchers and graduate students who know calculus, linear algebra, probability and basic Python.
Generative AI was used under human direction to produce manuscript text and companion code. The cover illustration was drawn by a script.