Human-in-the-Loop Machine Learning Active learning and annotation for human-centered AI 1st Edition By Robert (Munro) Monarch
$30.00
Secure checkout
Instant digital download
PDF document
Document details
- Pages
- 426
- File size
- 8.81 MB
- Format
- Digital PDF
- Category
- eBook[PDF]
Sign in or create a free account to continue. Your purchase will be saved in My Downloads.
About this ebook
Human-in-the-Loop Machine Learning: Active Learning and Annotation for Human-Centered AI (1st Edition) by Robert (Munro) Monarch is a highly practical, definitive guide published by Manning Publications that teaches data scientists how to optimize machine learning systems by effectively combining human intelligence with algorithms. [1, 2]
The book is available through major platforms like Amazon, O'Reilly Media, and Google Books. [1, 2, 3]
Core Focus & Overview
While most machine learning curricula focus purely on coding algorithms, this text addresses the real-world reality that data scientists spend the majority of their time managing, cleaning, and labeling data. It bridges the gap between Human-Computer Interaction (HCI) and artificial intelligence. Throughout the book, the author uses a real-world case study—classifying text messages from disaster response scenarios—to ground advanced concepts in practical engineering. [1, 2]
Core Book Structure
The textbook is divided into four main functional areas: [1]
- Introduction: Foundations of Human-in-the-Loop (HITL) architecture and initial system setups. [1]
- Active Learning: Deep dives into Uncertainty Sampling (identifying what the model is confused about) and Diversity Sampling (ensuring data covers all real-world scenarios), alongside multi-task active learning. [1, 2, 3, 4]
- Annotation: Managing human annotators, establishing rigorous Quality Control metrics, mitigating annotator bias, and performing data augmentation. [1, 2]
- Human-Computer Interaction: Designing optimal user interfaces for annotation tools and deploying complete human-centric AI products. [1]
Key Concepts Covered
- Smart Data Selection: Algorithms to calculate exactly which data points require human verification, drastically reducing manual labeling costs. [1, 2]
- Annotation Quality Control: Statistical methods to resolve disagreements among data labelers and evaluate worker accuracy. [1, 2]
- Advanced Workflows: Seamlessly blending transfer learning, self-supervision, and active learning within the same pipeline. [1]
- Multimodal Pipelines: Practical code and design patterns for building data for diverse tasks including text labeling, object detection, and semantic segmentation. [1]
About the Author
Dr. Robert Munro Monarch is a data scientist and engineer with a PhD from Stanford University focused on HITL machine learning for healthcare and disaster response. He has scaled massive data annotation and machine learning infrastructure for tech giants such as Amazon, Google, IBM, and Apple. [1]
File included
tmpphpyjMgUJ
8.81 MB
Related ebooks
