Description

Designed as a jumping-off point for engineers and decision-makers, this book provides a broad view of machine learning safety. It gives a validation road-map for developers and users of safety-critical systems where human lives are at stake. It addresses the limitations of machine learning systems at every stage of their lifecycle, and provides an overview of techniques to mitigate risks.

Chapters are structured to ease understanding of concepts: For each stage in the machine learning lifecycle, the reader is given a brief overview, list of terms, example methods, application methods, and resources for further investigation.

- Specifically addresses safety-critical systems where human users are at risk, physical property can be damaged, or financial losses are possible

- Covers datasets, robustness, reliability, interpretability, explainability, verification, validation, and operational safety

- Covers relevant safety standards

- Easy-to-digest overview of critical and fast-moving field

- Extensive bibliography

Those working in safety-critical areas, such as autonomous transportation, medical diagnostics and treatments, robotics and manufacturing, and financial systems, will find this book valuable.