en
Justin Bozonier

Test-Driven Machine Learning

Giv mig besked når bogen er tilgængelig
Denne bog er ikke tilgængelig i streaming pt. men du kan uploade din egen epub- eller fb2-fil og læse den sammen med dine andre bøger på Bookmate. Hvordan overfører jeg en bog?
Control your machine learning algorithms using test-driven development to achieve quantifiable milestonesAbout This BookBuild smart extensions to pre-existing features at work that can help maximize their valueQuantify your models to drive real improvementTake your knowledge of basic concepts, such as linear regression and Naive Bayes classification, to the next level and productionalize their modelsPlay what-if games with your models and techniques by following the test-driven exploration processWho This Book Is ForThis book is intended for data technologists (scientists, analysts, or developers) with previous machine learning experience who are also comfortable reading code in Python. You may be starting, or have already started, a machine learning project at work and are looking for a way to deliver results quickly to enable rapid iteration and improvement. Those looking for examples of how to isolate issues in models and improve them will find ideas in this book to move forward.What You Will LearnGet started with an introduction to test-driven development and familiarize yourself with how to apply these concepts to machine learningBuild and test a neural network deterministically, and learn to look for niche cases that cause odd model behaviourLearn to use the multi-armed bandit algorithm to make optimal choices in the face of an enormous amount of uncertaintyGenerate complex and simple random data to create a wide variety of test cases that can be codified into testsDevelop models iteratively, even when using a third-party libraryQuantify model quality to enable collaboration and rapid iterationAdopt simpler approaches to common machine learning algorithmsTake behaviour-driven development principles to articulate test intentIn DetailMachine learning is the process of teaching machines to remember data patterns, using them to predict future outcomes, and offering choices that would appeal to individuals based on their past preferences.Machine learning is applicable to a lot of what you do every day. As a result, you can't take forever to deliver your first iteration of software. Learning to build machine learning algorithms within a controlled test framework will speed up your time to deliver, quantify quality expectations with your clients, and enable rapid iteration and collaboration.This book will show you how to quantifiably test machine learning algorithms. The very different, foundational approach of this book starts every example algorithm with the simplest thing that could possibly work. With this approach, seasoned veterans will find simpler approaches to beginning a machine learning algorithm. You will learn how to iterate on these algorithms to enable rapid delivery and improve performance expectations.The book begins with an introduction to test driving machine learning and quantifying model quality. From there, you will test a neural network, predict values with regression, and build upon regression techniques with logistic regression. You will discover how to test different approaches to naive bayes and compare them quantitatively, along with how to apply OOP (Object-Oriented Programming) and OOP patterns to test-driven code, leveraging SciKit-Learn.Finally, you will walk through the development of an algorithm which maximizes the expected value of profit for a marketing campaign by combining one of the classifiers covered with the multiple regression example in the book.Style and approachAn example-driven guide that builds a deeper knowledge and understanding of iterative machine learning development, test by test. Each topic develops solutions using failing tests to illustrate problems; these are followed by steps to pass the tests, simply and straightforwardly. Topics which use generated data explore how the data was generated, alongside explanations of the assumptions behind different machine learning techniques.
Denne bog er ikke tilgængelig i øjeblikket
253 trykte sider
Udgivelsesår
2015
Har du allerede læst den? Hvad synes du om den?
👍👎

På boghylderne

  • Ivan
    Tech
    • 14
    • 1
  • Mikkel Terpe Woods
    Mikkel
    • 8
fb2epub
Træk og slip dine filer (ikke mere end 5 ad gangen)