Book Image

Hands-On Machine Learning with scikit-learn and Scientific Python Toolkits

By : Tarek Amr
Book Image

Hands-On Machine Learning with scikit-learn and Scientific Python Toolkits

By: Tarek Amr

Overview of this book

Machine learning is applied everywhere, from business to research and academia, while scikit-learn is a versatile library that is popular among machine learning practitioners. This book serves as a practical guide for anyone looking to provide hands-on machine learning solutions with scikit-learn and Python toolkits. The book begins with an explanation of machine learning concepts and fundamentals, and strikes a balance between theoretical concepts and their applications. Each chapter covers a different set of algorithms, and shows you how to use them to solve real-life problems. You’ll also learn about various key supervised and unsupervised machine learning algorithms using practical examples. Whether it is an instance-based learning algorithm, Bayesian estimation, a deep neural network, a tree-based ensemble, or a recommendation system, you’ll gain a thorough understanding of its theory and learn when to apply it. As you advance, you’ll learn how to deal with unlabeled data and when to use different clustering and anomaly detection algorithms. By the end of this machine learning book, you’ll have learned how to take a data-driven approach to provide end-to-end machine learning solutions. You’ll also have discovered how to formulate the problem at hand, prepare required data, and evaluate and deploy models in production.
Table of Contents (18 chapters)
1
Section 1: Supervised Learning
8
Section 2: Advanced Supervised Learning
13
Section 3: Unsupervised Learning and More

Getting to know additional linear classifiers

Before ending this chapter, it is useful to highlight some additional linear classification algorithms:

  • SGD is a versatile solver. As mentioned earlier, it can perform a logistic regression classification in addition to SVM and perceptron classification, depending on the loss function used. It also allows regularized penalties.
  • The rideclassifier converts class labels into 1 and -1 and treats the problem as a regression task. It also deals well with non-binary classification tasks. Due to its design, it uses a different set of solvers, so it's worth trying as it may be quicker to learn when dealing with a large number of classes.
  • LinearSupportVectorClassification (LinearSVC) is another linear model. Rather than log loss, it uses the hinge function, which aims to find class boundaries where the samples of each class are as far as possible from the boundaries. This is not to be confused with...