Book Image

Mastering Azure Machine Learning

By : Christoph Körner, Kaijisse Waaijer
Book Image

Mastering Azure Machine Learning

By: Christoph Körner, Kaijisse Waaijer

Overview of this book

The increase being seen in data volume today requires distributed systems, powerful algorithms, and scalable cloud infrastructure to compute insights and train and deploy machine learning (ML) models. This book will help you improve your knowledge of building ML models using Azure and end-to-end ML pipelines on the cloud. The book starts with an overview of an end-to-end ML project and a guide on how to choose the right Azure service for different ML tasks. It then focuses on Azure Machine Learning and takes you through the process of data experimentation, data preparation, and feature engineering using Azure Machine Learning and Python. You'll learn advanced feature extraction techniques using natural language processing (NLP), classical ML techniques, and the secrets of both a great recommendation engine and a performant computer vision model using deep learning methods. You'll also explore how to train, optimize, and tune models using Azure Automated Machine Learning and HyperDrive, and perform distributed training on Azure. Then, you'll learn different deployment and monitoring techniques using Azure Kubernetes Services with Azure Machine Learning, along with the basics of MLOps—DevOps for ML to automate your ML process as CI/CD pipeline. By the end of this book, you'll have mastered Azure Machine Learning and be able to confidently design, build and operate scalable ML pipelines in Azure.
Table of Contents (20 chapters)
1
Section 1: Azure Machine Learning
4
Section 2: Experimentation and Data Preparation
9
Section 3: Training Machine Learning Models
15
Section 4: Optimization and Deployment of Machine Learning Models
19
Index

Ensuring reproducible builds and deployments

DevOps has many different meanings, but it is usually oriented toward enabling rapid and high-quality deployments when source code changes. One way of achieving high-quality operational code is to guarantee reproducible and predictable builds, which is also crucial for creating reproducible ML pipelines. While it seems obvious for application development that the compiled binary will look and behave in a similar manner, with only a few minor configuration changes, the same is not true for the development of ML pipelines.

There are four main problems that ML engineers and data scientists face that make building reproducible deployments very difficult:

  • The development process is often performed in notebooks, so it is not always linear.
  • There are mismatching library versions and drivers.
  • Source data can be changed or modified.
  • Non-deterministic optimization techniques can lead to completely different outputs...