Launch of Conformal Prediction Book 🦫
A Short Guide For Quantifying Uncertainty Of Machine Learning Models
The ebook “Introduction To Conformal Prediction With Python” is finally published 🥳🥳🥳
tl;dr you can get the book at early bird price with 50% discount
The last 2½ months have been … intense. The only thing I spent my time with was conformal prediction. I dreamt about it at least twice. And a good part of the time I worked from my couch after knee surgery.
I’m happy the project is finished (and that my knee is much better). At the same time, I’m grateful that my job as a full-time writer allows me to spend so much time learning about new topics. And that’s only possible because of readers like you! A big thank you at this point ❣️
Why Conformal Prediction?
My first encounter with conformal prediction was years ago. Back then, I didn’t understand it, I didn’t like it, and anyways I was looking for something else and I moved on.
But years later, I gave conformal prediction another try. A bit more serious this time. But also a lot had happened in the meantime in terms of research and material in just these few years.
This time, when I looked closer at conformal prediction, I found a method for uncertainty quantification that makes a lot of sense and is just very practical.
So I decided to double down on the topic and created an e-mail course that you might have participated in. And based on this e-mail course I wrote an introduction book that I publish today. 🥳
After this intense time learning about conformal prediction, I’m still convinced that it’s a no-brainer to add it to your modeling toolbox:
You can combine conformal prediction with any uncertainty quantification method that you already use. Using Bayesian models? You can use CP on top. No obligation to use CP exclusively.
It’s model-agnostic. Replacing an SVM for an xgboost model? With conformal prediction, you don’t have to change the way you quantify uncertainty. Meaning no changes in the user interface and no need to explain to your coworkers anew how to interpret the uncertainty measures.
But the best part of conformal prediction: it comes with a coverage guarantee of the true outcome. And this guarantee is comparably “cheap” since the only requirement is that calibration and new data are exchangeable. No priors. No distribution assumptions.
So it’s really a lightweight and easy-to-use addition to your modeling setup. But given all these arguments in favor of conformal prediction why are not more people using it?
My take: Conformal prediction is still relatively new and predominantly “lives” in academia. Conformal prediction is becoming more widespread and more accessible, but educational coverage — especially entry-level educational material — is nowhere near that of methods like bootstrapping.
Why This Book?
To bring conformal prediction to the streets.
While conformal prediction is easy-to-use and practical, learning about it means being confronted with the science and math behind it. “Introduction To Conformal Prediction With Python” approaches conformal prediction from an intuitive and practical angle. The book:
focuses more on intuition than math,
makes it easy to get started with conformal prediction,
provides practical code examples,
enables you to understand more complex conformal prediction algorithms,
is short, and easy to read.
You can get the ebook directly from Leanpub with a 50% discount. Only this week.
The purchase is without risk because there’s a 60 days guarantee to return it (whatever that means for an ebook 🤷♂️).
What Others Said About The Book
Great practical examples, easy explanations, and highly entertaining. If you want to learn about the best Uncertainty Quantification framework for the 21st century, don't miss out on this book.
Valeriy Manokhin, Managing Director at Open Predictive Technologies & Creator of Awesome Conformal Prediction
This concise book is accessible, lucid, and full of helpful code snippets. It explains the mathematical ideas with clarity and provides the reader with practical examples that illustrate the essence of conformal prediction, a powerful idea for uncertainty quantification.
Junaid Butt, Research Software Engineer at IBM Research
Modern statistics can be a difficult topic, but Christoph has managed to make it feel easy, practical, and fun! Reading this book is a great first step towards gaining mastery of conformal prediction and related topics.
Anastasios Angelopoulos, Researcher at the University of California, Berkeley