Chapter 1, Understanding the End-to-End Machine Learning Process, covers the history of ML, the scenarios in which to apply ML, the statistical knowledge necessary, and the steps and components required for running a custom end-to-end ML project. Its purpose is to bring every reader to the same foundational level. Due to that, some sections might be a recap for readers that are very knowledgeable about ML but still might hold some useful practical tips and guidelines for them. It is also designed to be the guide for the rest of the book, where every step in the ML process will point to the chapters covering them in detail.

Chapter 2, Choosing the Right Machine Learning Service in Azure, helps us understand and classify the available Azure services for ML. We will define the scenarios in which to use certain services and we will conclude that for building custom ML models, Azure Machine Learning is the best choice. From this chapter onward, we use the available tooling in the Azure Machine Learning service to perform all upcoming tasks in the ML process.

Chapter 3, Preparing the Azure Machine Learning Workspace, covers the setup of the Azure Machine Learning service and some initial hands-on ML training using the service. We will perform ML training experiments while learning how to track the experiments, plot metrics, and create snapshots of ML runs with the available tooling in Azure Machine Learning.

Chapter 4, Ingesting Data and Managing Datasets, covers the available Azure services to store our underlying data and how to set them up in Azure. Furthermore, we will understand how we can bring the required data to these services either manually or automatically through Extract, Transform, and Load (ETL) processes and how we can integrate other Azure data services with Azure Machine Learning. Finally, we will introduce the concepts of datastores and datasets in Azure Machine Learning and how to use them in our experiment runs.

Chapter 5, Performing Data Analysis and Visualization, covers the steps required to explore and preprocess an ML dataset. We will understand the difference between a tabular and a file dataset, and we will learn how to clean our dataset, correlate features, and use statistical properties and domain knowledge to get insight into our dataset. Using what we've learned, we will go hands-on on a real-life dataset to apply our knowledge. Finally, we will have a peek at some popular embedding techniques such as PCA, LDA, t-SNE, and UMAP.

Chapter 6, Feature Engineering and Labeling, covers the important process of creating or adapting features in our dataset and creating labels for supervised ML training. We will understand the reasons for changing our features and we will glance at a variety of available methods to create, transform, extract, and select features in a dataset, which we will then use on our real-life dataset. Furthermore, we will explore techniques to label different types of datasets and go hands-on with the Data Labeling tool in Azure Machine Learning.

Chapter 7, Advanced Feature Extraction with NLP, takes us one step further to extract features from textual and categorical data – a problem that users are faced with often when training ML models. This chapter will describe the foundations of feature extraction for Natural Language Processing (NLP). This will help us to create semantic embeddings from categorical and textual data using techniques including n-grams, Bag of Words, TF-IDF, Word2Vec, and more.

Chapter 8, Azure Machine Learning Pipelines, covers how we can incorporate what we have learned in an automated preprocessing and training pipeline using Azure Machine Learning pipelines. We will learn how to split our code into modular pipeline steps and how to parameterize and trigger pipelines through endpoints and scheduling. Finally, we will build a couple of training pipelines and learn how to integrate them into other Azure services.

Chapter 9, Building ML Models Using Azure Machine Learning, teaches you how to use ensembling techniques to build a traditional ML model in Azure. This chapter focuses on decision tree-based ensemble learning with popular state-of-the-art boosting and bagging techniques using LightGBM in Azure Machine Learning. This will help you to apply concepts of bagging and boosting on ML models.

Chapter 10, Training Deep Neural Networks on Azure, covers training more complex parametric models using deep learning for better generalization over large datasets. We will give a short and practical overview of which situations deep learning can be applied well to and how it differs from the more traditional ML approaches. After that, we will discuss rational and practical guidelines to finally train a Convolutional Neural Network (CNN) on Azure Machine Learning using Keras.

Chapter 11, Hyperparameter Tuning and Automated Machine Learning, covers the optimization of the ML training process and how to automate it to avoid human errors. These tuning tricks will help you to train models faster and more efficiently. Therefore, we will look at hyperparameter tuning (also called HyperDrive in Azure Machine Learning), a standard technique for optimizing all external parameters of an ML model. By evaluating different sampling techniques for hyperparameter tuning, such as random sampling, grid sampling, and Bayesian optimization, you will learn how to efficiently manage the trade-offs between runtime and model performance. Then, we will generalize from hyperparameter optimization to automating the complete end-to-end ML training process using Azure automated machine learning.

Chapter 12, Distributed Machine Learning on Azure, looks into distributed and parallel computing algorithms and frameworks for efficiently training ML models in parallel on GPUs. The goal of this chapter is to build an environment in Azure where you can speed up the training process of classical ML and deep learning models by adding more machines to your training environment and hence scaling out the cluster.

Chapter 13, Building a Recommendation Engine in Azure, dives into traditional and modern recommendation engines that often combine the technologies and techniques covered in the previous chapters. We will take a quick look at the different types of recommendation engines, what data is needed for each type, and what can be recommended using these different approaches, such as content-based recommendations and rating-based recommendation engines. We will combine both techniques into a single hybrid recommender and learn about state-of-the-art techniques for modern recommendation engines.

Chapter 14, Model Deployment, Endpoints, and Operations, finally covers how to bring our ML models into a production environment, by deploying them either to a batch cluster for offline scoring or as an endpoint for online scoring. To achieve that, we are going to package the model and execution runtime, register both in a model registry, and deploy them to an execution environment. We will auto-deploy models from Azure Machine Learning to Azure Kubernetes Service with only a few lines of code. Finally, you will learn how to monitor your target environments using out-of-the-box custom metrics.

Chapter 15, Model Interoperability, Hardware Optimization, and Integrations, covers methods to standardize deployment model formats using the Open Neural Network eXchange (ONNX), what Field Programmable Gate Arrays (FPGA) are, and how to use them as a deployment target in Azure. Further, we will learn how to integrate Azure Machine Learning with other Microsoft services such as Azure IoT Edge and Power BI. Here, we will understand the fundamental differences between FPGAs and GPUs in terms of performance, cost, and efficiency and we will go hands-on in Power BI to integrate one of our previously deployed endpoints.

Chapter 16, Bringing Models into Production with MLOps, finally covers how we put data ingestion, data preparation, our ML training and deployment pipelines, and any required script into one end-to-end operation. This includes the creation of environments; starting, stopping, and scaling clusters; submitting experiments; performing parameter optimization; and deploying full-fledged scoring services on Kubernetes. We will reuse all the concepts we applied previously to build a version-controlled, reproducible, automated ML training and deployment process as a Continuous Integration/Continuous Deployment (CI/CD) pipeline in Azure DevOps.

Chapter 17, Preparing for a Successful ML Journey, ends the book by giving you a summary of the major concepts we learned throughout it and highlights what really matters when performing ML. We reiterate the importance of a clean base infrastructure, monitoring, and automation and discuss the ever-changing nature of ML and cloud-based services. Finally, we cover one of the most important topics, which we glanced over throughout the book, ethics in data processing. We will discuss your responsibility to have fair and explainable ML models and how Azure Machine Learning and open source tooling can help you achieve that.