UW’s Kotthoff and Students Co-Publish Book on Applied Machine Learning

In an era when data drives decisions, mastering machine learning has become essential across industries and academia. From making better business decisions to developing advanced materials, machine learning and data science are crucial, and understanding them is important to the success of University of Wyoming graduates.

But getting started can be daunting, especially for those looking to apply machine learning in their disciplines.

A new book that focuses on applying machine learning aims to help students, researchers and practitioners. Lars Kotthoff, an associate professor in UW’s Department of Electrical Engineering and Computer Science, co-edited and co-wrote the book with a team of international researchers based at various institutions in Europe, primarily Ludwig Maximilian University of Munich in Germany. Authors also include UW students Natalie Foss, of Colorado Springs, Colo., who graduated with a bachelor’s degree in computer science from UW in 2023, and Damir Pulatov, of Tashkent, Uzbekistan, who is currently finishing his Ph.D. in computer science.

The book provides a comprehensive introduction to machine learning in R, a programming language widely used in the sciences. It contains many examples that readers can run directly and adapt for their needs, enabling even machine learning novices to get started immediately.

The later chapters cover more advanced topics, ranging from methods for getting the best performance out of your model through hyperparameter optimization, to building complex machine learning pipelines, to working with geospatial data. Technical documentation on how to extend the software, such as with custom modeling approaches, and how to run it on high-performance computing infrastructure also is included.

“This work represents the culmination of more than a decade of work to implement the software and comprehensively document it,” says Kotthoff, who also is a founding adjunct faculty member in UW’s School of Computing and a Presidential Faculty Fellow.

The open-source mlr3 machine learning ecosystem, on which the book is based, has been used widely in academia and industry for many years. Kotthoff is one of the principal developers of the software, which also includes contributions by his students.

“The mlr3 ecosystem makes common machine learning tasks, such as comparing different modeling approaches, very easy through convenience functions that provide a lot of functionality in very little code,” he says. “The focus is very much on the end user, who does not need to be a machine learning expert, and empowering them to use state-of-the-art machine learning for their problems."

The book, published by Chapman and Hall, is geared toward both machine learning novices and advanced users. It is available in its entirety for free online at https://mlr3book.mlr-org.com/.

“Academic textbooks can be very expensive, which represents a significant barrier for students, especially from low-income backgrounds,” Kotthoff says. “We are very happy that our publisher generously allowed us to keep the book online for free, improving access to these important concepts.”

Development of both the book and the mlr3 software continues.

“We will continue to improve and release new versions of mlr3 and the book,” Kotthoff says. “Machine learning is a fast-moving field. Advances are made regularly, and we are committed to making these available to users of mlr3.”

The mlr3 ecosystem is actively developed on Github at https://github.com/mlr-org/, with different packages that provide different functionality and dozens of contributors. The software is extensively tested and extensively documented.

For more information about the book or the mlr3 software, email Kotthoff at larsko@uwyo.edu.