Foreword

Machine Learning (ML) is one of the most impactful technologies of the last 10 years, fueled by the exponential growth of electronic data about people and their interaction with the world and each other, as well as the availability of massive computing power to extract patterns from data. Applications of ML are already affecting all of us in everyday life, whether it's face recognition in modern cameras, personalized web or product searches, or even the detection of road sign patterns in modern cars. Machine learning is a set of algorithms that learn prediction programs from past data in order to use them for future predictions—whether the prediction programs are represented as decision trees, as neural networks, or via nearest-neighbor functions.

Another influential development in computer science is the invention of F#. Less than 10 years ago, functional programming was a more of an academic endeavor than a style of programming and software development used in production systems. The development of F# since 2005 changed this forever. With F#, programmers are not only able to benefit from type inference and easy parallelization of workflows, but they also get the runtime performance that they are used to from programming in other .NET languages, such as C#. I personally witnessed this transformation at Microsoft Research and saw how data-intensive applications could be written much more safely in less than 100 lines of F# code compared to thousands of lines of C# code.

A critically important ingredient of ML is data; it's the lifeblood of any ML algorithm. Parsing, cleaning, and visualizing data is the basis of any successful ML application and constitutes the majority of the time that practitioners spend in making machine learning systems work. F# proves to be the perfect bridge between data processing and analysis, with ML on one hand and the ability to invent new ML algorithms on the other hand.

In this book, Sudipta Mukherjee introduces the reader to the basics of machine learning, ranging from supervised methods, such as classification learning and regression, to unsupervised methods, such as K-means clustering. Sudipta focuses on the applied aspects of machine learning and develops all algorithms in F#, both natively as well as by integrating with .NET libraries such as WekaSharp, Accord.Net and Math.Net. He covers a wide range of algorithms for classification and regression learning and also explores more novel ML concepts, such as anomaly detection. The book is enriched with directly applicable source code examples, and the reader will enjoy learning about modern machine learning algorithms through the numerous examples provided.

Dr. Ralf Herbrich

Director of Machine Learning Science at Amazon