Mastering Azure Machine Learning by Körner, Christoph -- Read -- Imperial Library of Trantor

Index

Title Page Copyright and Credits

Mastering Azure Machine Learning

About Packt

Why subscribe?

Contributors

About the authors About the reviewers Packt is searching for authors like you

Preface

Who this book is for What this book covers To get the most out of this book

Download the example code files Download the color images Conventions used

Get in touch

Reviews

Section 1: Azure Machine Learning Services Building an End-To-End Machine Learning Pipeline in Azure

Performing descriptive data exploration

Moving data to the cloud Understanding missing values Visualizing data distributions Finding correlated dimensions Measuring feature and target dependencies for regression Visualizing feature and label dependency for classification

Exploring common techniques for data preparation

Labeling the training data Normalization and transformation in machine learning Encoding categorical variables A feature engineering example using time-series data Using NLP to extract complex features from text

Choosing the right ML model to train data

Choosing an error metric The training and testing split Achieving great performance using tree-based ensemble models Modeling large and complex data using deep learning techniques

Optimization techniques

Hyperparameter optimization Model stacking AutoML

Deploying and operating models

Batch scoring using pipelines Real-time scoring using a container-based web service Tracking model performance, telemetry, and data skew

Summary

Choosing a Machine Learning Service in Azure

Demystifying the different Azure services for ML

Choosing an Azure service for ML Choosing a compute target for an Azure ML service

Azure Cognitive Services and Custom Vision

Azure Cognitive Services Custom Vision—customizing the Cognitive Services API

Azure ML tools with GUIs

Azure ML Studio (classic) Azure Automated ML Microsoft Power BI

The Azure ML service

Organizing experiments and models in Azure ML Deployments through Azure ML

Summary

Section 2: Experimentation and Data Preparation Data Experimentation and Visualization Using Azure

Preparing your Azure ML workspace

Setting up the ML Service workspace Running a simple experiment with Azure ML Logging metrics and tracking results Scheduling and running scripts Adding cloud compute to the workspace

Visualizing high-dimensional data

Tracking figures in experiments in Azure ML Unsupervised dimensionality reduction with PCA Using LDA for supervised projections Non-linear dimension reduction with t-SNE Generalizing t-SNE with UMAP

Summary

ETL, Data Preparation, and Feature Extraction

Managing data and dataset pipelines in the cloud

Getting data into the cloud

Organizing data in data stores and datasets

Managing data in Azure ML

Versioning datasets and dataset definitions Taking data snapshots for reproducibility The life cycle of a dataset

Exploring data registered in the Azure ML service

Exploring the datasets Exploring the data

Preprocessing and feature engineering with Azure ML DataPrep

Parsing different data formats

Loading delimiter-separated data Parsing JSON data Loading binary column-store data in Parquet format

Building a data transformation pipeline in Azure ML

Generating features through expression Data type conversions Deriving columns by example Imputing missing values Label and one-hot encoding Transformations and scaling Filtering columns and rows

Writing the processed data back to a dataset

Summary

Advanced Feature Extraction with NLP

Understanding categorical data

Comparing textual, categorical, and ordinal data Transforming categories into numeric values

Orthogonal embedding using one-hot encoding

Categories versus text

Building a simple bag-of-words model

A naive bag-of-words model using counting Tokenization – turning a string into a list of words Stemming – rule-based removal of affixes Lemmatization – dictionary-based word normalization A bag-of-words model in scikit-learn

Leveraging term importance and semantics

Generalizing words using n-grams and skip-grams Reducing word dictionary size using SVD Measuring the importance of words using tf-idf Extracting semantics using word embeddings

Implementing end-to-end language models

End-to-end learning of token sequences State-of-the-art sequence-to-sequence models Text analytics using Azure Cognitive Services

Summary

Section 3: Training Machine Learning Models Building ML Models Using Azure Machine Learning

Working with tree-based ensemble classifiers

Understanding a simple decision tree

Advantages of a decision tree Disadvantages of a decision tree

Combining classifiers with bagging Optimizing classifiers with boosting rounds

Training an ensemble classifier model using LightGBM

LightGBM in a nutshell Preparing the data Setting up the compute cluster and execution environment Building a LightGBM classifier Scheduling the training script on the Azure ML cluster

Summary

Training Deep Neural Networks on Azure

Introduction to deep learning

Why DL? From neural networks to DL Comparing classical ML and DL

Training a CNN for image classification

Training a CNN from scratch in your notebook Generating more input data using augmentation Moving training to a GPU cluster using Azure ML compute Improving your performance through transfer learning

Summary

Hyperparameter Tuning and Automated Machine Learning

Hyperparameter tuning to find the optimal parameters

Sampling all possible parameter combinations using grid search Trying random combinations using random search Converging faster using early termination

The median stopping policy The truncation selection policy The bandit policy A HyperDrive configuration with termination policy

Optimizing parameter choices using Bayesian optimization

Finding the optimal model with AutoML

Advantages and benefits of AutoML A classification example

Summary

Distributed Machine Learning on Azure ML Clusters

Exploring methods for distributed ML

Training independent models on small data in parallel Training a model ensemble on large datasets in parallel Fundamental building blocks for distributed ML Speeding up DL with data-parallel training Training large models with model-parallel training

Using distributed ML in Azure

Horovod—a distributed DL training framework Implementing the HorovodRunner API for a Spark job Running Horovod on Azure ML compute

Summary

Building a Recommendation Engine in Azure

Introduction to recommender engines Content-based recommendations

Measuring similarity between items Feature engineering for content-based recommenders Content-based recommendations using gradient boosted trees

Collaborative filtering—a rating-based recommendation engine

What is a rating? Explicit feedback as opposed to implicit feedback Predicting the missing ratings to make a recommendation Scalable recommendations using ALS factorization

Combining content and ratings in hybrid recommendation engines

Building a state-of-the-art recommender using the Matchbox Recommender

Automatic optimization through reinforcement learning

An example using Azure Personalizer in Python

Summary

Section 4: Optimization and Deployment of Machine Learning Models Deploying and Operating Machine Learning Models

Deploying ML models in Azure

Understanding the components of an ML model Registering your models in a model registry Customizing your deployment environment Choosing a deployment target in Azure

Building a real-time scoring service Implementing a batch scoring pipeline Inference optimizations and alternative deployment targets

Profiling models for optimal resource configuration Portable scoring through the ONNX runtime Fast inference using FPGAs in Azure Alternative deployment targets

Monitoring Azure ML deployments

Collecting logs and infrastructure metrics Tracking telemetry and application metrics

Summary

MLOps - DevOps for Machine Learning

Ensuring reproducible builds and deployments

Version-controlling your code Registering snapshots of your data Tracking your model metadata and artifacts Scripting your environments and deployments

Validating your code, data, and models

Rethinking unit testing for data quality Integration testing for ML End-to-end testing using Azure ML Continuous profiling of your model

Summary

What's Next?

Understanding the importance of data The future of ML is automated Change is the only constant – preparing for change Focusing first on infrastructure and monitoring Controlled rollouts and A/B testing Summary

Other Books You May Enjoy

Leave a review - let other readers know what you think

← Prev
Back
Next →

← Prev
Back
Next →