Reviews
AI technology advances day by day. As AI has reached performance approaching human levels, a new research field has emerged: AI safety. AI whose workings are hard to understand can deliver astonishing results, and unfounded rumors keep spreading. Separate from that, research into safe AI is clearly necessary. This book addresses ways to develop and deploy AI safely. Part one, titled The Theory and Practice of AI Risk Management, introduces the current state of AI regulation, interpretable and
Machine Learning Risk Management with Python teaches readers to understand the AI model risk management process and how to apply it in practice. It introduces current tools and techniques for training explainable AI models and managing safety, bias, security, and privacy issues. Readers can learn practical strategies for operating ML systems responsibly and developing them sustainably. The book provides basic guidance on existing standards, legal requirements, and evaluation methods for
Why explainability matters when choosing a model: The main factors in choosing a suitable machine learning model include the type of problem, such as classification or regression; each model's accuracy; the size and amount of data; and business costs and time. Citing a paper by Cynthia Rudin, the book argues at the outset that it is wrong to assume there must be a tradeoff between accuracy and interpretability, and explains why explainable models are important and necessary. This matters more
Patrick Hall et al.: Machine Learning Risk Management with Python
Like any technology, machine learning can fail through unintended misuse or deliberate abuse. To realize the real benefits of this fascinating technology, organizations and the public need to mitigate those risk factors. The initial standards this book seeks to follow have only recently begun to take shape, and machine learning practice still lacks widely recognized professional certifications or best practices. As a result, practitioners bear responsibility for the outcomes when machine learning technologies are deployed. This book offers practitioners new ways to understand AI model risk management, train explainable models with common Python tools, and manage safety, bias, security, and privacy issues. Taking a nuanced approach to reducing ML risks, it gives readers a valuable resource for successfully deploying ML systems responsibly and sustainably. Key topics include understanding technical approaches to responsible AI, learning how to build successful and impactful AI risk management practices, reviewing basic guidance on existing standards, laws, and evaluations for adopting AI technology, exploring the new NIST AI Risk Management Framework, and using interactive resources on GitHub and Colab.
Korean title: 머신러닝 리스크 관리 with 파이썬
Korean author: 패트릭 홀 외
Korean publisher: 한빛미디어
ISBN-13: 9791169212380
Reviews
AI technology advances day by day. As AI has reached performance approaching human levels, a new research field has emerged: AI safety. AI whose workings are hard to understand can deliver astonishing results, and unfounded rumors keep spreading. Separate from that, research into safe AI is clearly necessary. This book addresses ways to develop and deploy AI safely. Part one, titled The Theory and Practice of AI Risk Management, introduces the current state of AI regulation, interpretable and
Machine Learning Risk Management with Python teaches readers to understand the AI model risk management process and how to apply it in practice. It introduces current tools and techniques for training explainable AI models and managing safety, bias, security, and privacy issues. Readers can learn practical strategies for operating ML systems responsibly and developing them sustainably. The book provides basic guidance on existing standards, legal requirements, and evaluation methods for
Why explainability matters when choosing a model: The main factors in choosing a suitable machine learning model include the type of problem, such as classification or regression; each model's accuracy; the size and amount of data; and business costs and time. Citing a paper by Cynthia Rudin, the book argues at the outset that it is wrong to assume there must be a tradeoff between accuracy and interpretability, and explains why explainable models are important and necessary. This matters more
