A principle-driven textbook that teaches students and practitioners to reason quantitatively about machine learning systems, from data pipelines to deployment.Machine learning has crossed from research into engineering practice, yet the field lacks a comprehensive treatment of principles, vocabulary, and quantitative reasoning tools. Filling that gap, this innovative textbook treats machine learning systems not as a collection of tools and frameworks, but as an engineering discipline governed by physical constraints. Introduction to Machine Learning Systems develops quantitative frameworks that decompose system performance into measurable components, giving readers the ability to diagnose bottlenecks, predict trade-offs, and design systems that work—by reasoning from first principles, not recipes.    Organized in four parts—Foundations, Build, Optimize, and Deploy—the book covers the complete ML systems lifecycle: data engineering, neural network computation and architectures, framework internals, training infrastructure, data selection, model compression, hardware acceleration, benchmarking, serving systems, ML operations, and responsible engineering including fairness, privacy, security, and sustainability. The scope encompasses systems from embedded devices to cloud-based accelerators on a single compute node, the fundamental unit of ML computation and the prerequisite for everything built on top of it.  Develops quantitative reasoning tools that let readers diagnose system bottlenecks and predict trade-offs  Covers the full ML systems lifecycle end-to-end, from data pipelines through training, optimization, deployment, and operations Teaches enduring principles rather than current tools Treats fairness, privacy, security, and environmental sustainability as engineering problems with measurable solutions Features rich pedagogy including learning objectives, self-check questions, worked calculations, and real-world production failure case studies Is based on the author's popular Harvard course and the TinyML edX program  Offers interactive labs, lecture slides, and the companion TinyTorch educational framework

Shrinking feature size and diminishing supply voltage are making circuits sensitive to supply voltage fluctuations within the microprocessor, caused by normal workload activity changes. If left unattended, voltage fluctuations can lead to timing violations or even transistor lifetime issues that degrade processor robustness. Mechanisms that learn to tolerate, avoid, and eliminate voltage fluctuations based on program and microarchitectural events can help steer the processor clear of danger, thus enabling tighter voltage margins that improve performance or lower power consumption. We describe the problem of voltage variation and the factors that influence this variation during processor design and operation. We also describe a variety of runtime hardware and software mitigation techniques that either tolerate, avoid, and/or eliminate voltage violations. We hope processor architects will find the information useful since tolerance, avoidance, and elimination are generalizable constructs that can serve as a basis for addressing other reliability challenges as well. Table of Contents: Introduction / Modeling Voltage Variation / Understanding the Characteristics of Voltage Variation / Traditional Solutions and Emerging Solution Forecast / Allowing and Tolerating Voltage Emergencies / Predicting and Avoiding Voltage Emergencies / Eliminiating Recurring Voltage Emergencies / Future Directions on Resiliency

Shrinking feature size and diminishing supply voltage are making circuits sensitive to supply voltage fluctuations within the microprocessor, caused by normal workload activity changes. If left unattended, voltage fluctuations can lead to timing violations or even transistor lifetime issues that degrade processor robustness. Mechanisms that learn to tolerate, avoid, and eliminate voltage fluctuations based on program and microarchitectural events can help steer the processor clear of danger, thus enabling tighter voltage margins that improve performance or lower power consumption. We describe the problem of voltage variation and the factors that influence this variation during processor design and operation. We also describe a variety of runtime hardware and software mitigation techniques that either tolerate, avoid, and/or eliminate voltage violations. We hope processor architects will find the information useful since tolerance, avoidance, and elimination are generalizable constructsthat can serve as a basis for addressing other reliability challenges as well. Table of Contents: Introduction / Modeling Voltage Variation / Understanding the Characteristics of Voltage Variation / Traditional Solutions and Emerging Solution Forecast / Allowing and Tolerating Voltage Emergencies / Predicting and Avoiding Voltage Emergencies / Eliminiating Recurring Voltage Emergencies / Future Directions on Resiliency

A principle-driven textbook that teaches students and practitioners to reason quantitatively about machine learning systems, from data pipelines to deployment.Machine learning has crossed from research into engineering practice, yet the field lacks a comprehensive treatment of principles, vocabulary, and quantitative reasoning tools. Filling that gap, this innovative textbook treats machine learning systems not as a collection of tools and frameworks, but as an engineering discipline governed by physical constraints. Introduction to Machine Learning Systems develops quantitative frameworks that decompose system performance into measurable components, giving readers the ability to diagnose bottlenecks, predict trade-offs, and design systems that work—by reasoning from first principles, not recipes.    Organized in four parts—Foundations, Build, Optimize, and Deploy—the book covers the complete ML systems lifecycle: data engineering, neural network computation and architectures, framework internals, training infrastructure, data selection, model compression, hardware acceleration, benchmarking, serving systems, ML operations, and responsible engineering including fairness, privacy, security, and sustainability. The scope encompasses systems from embedded devices to cloud-based accelerators on a single compute node, the fundamental unit of ML computation and the prerequisite for everything built on top of it.  Develops quantitative reasoning tools that let readers diagnose system bottlenecks and predict trade-offs  Covers the full ML systems lifecycle end-to-end, from data pipelines through training, optimization, deployment, and operations Teaches enduring principles rather than current tools Treats fairness, privacy, security, and environmental sustainability as engineering problems with measurable solutions Features rich pedagogy including learning objectives, self-check questions, worked calculations, and real-world production failure case studies Is based on the author's popular Harvard course and the TinyML edX program  Offers interactive labs, lecture slides, and the companion TinyTorch educational framework

Shrinking feature size and diminishing supply voltage are making circuits sensitive to supply voltage fluctuations within the microprocessor, caused by normal workload activity changes. If left unattended, voltage fluctuations can lead to timing violations or even transistor lifetime issues that degrade processor robustness. Mechanisms that learn to tolerate, avoid, and eliminate voltage fluctuations based on program and microarchitectural events can help steer the processor clear of danger, thus enabling tighter voltage margins that improve performance or lower power consumption. We describe the problem of voltage variation and the factors that influence this variation during processor design and operation. We also describe a variety of runtime hardware and software mitigation techniques that either tolerate, avoid, and/or eliminate voltage violations. We hope processor architects will find the information useful since tolerance, avoidance, and elimination are generalizable constructs that can serve as a basis for addressing other reliability challenges as well. Table of Contents: Introduction / Modeling Voltage Variation / Understanding the Characteristics of Voltage Variation / Traditional Solutions and Emerging Solution Forecast / Allowing and Tolerating Voltage Emergencies / Predicting and Avoiding Voltage Emergencies / Eliminiating Recurring Voltage Emergencies / Future Directions on Resiliency

Shrinking feature size and diminishing supply voltage are making circuits sensitive to supply voltage fluctuations within the microprocessor, caused by normal workload activity changes. If left unattended, voltage fluctuations can lead to timing violations or even transistor lifetime issues that degrade processor robustness. Mechanisms that learn to tolerate, avoid, and eliminate voltage fluctuations based on program and microarchitectural events can help steer the processor clear of danger, thus enabling tighter voltage margins that improve performance or lower power consumption. We describe the problem of voltage variation and the factors that influence this variation during processor design and operation. We also describe a variety of runtime hardware and software mitigation techniques that either tolerate, avoid, and/or eliminate voltage violations. We hope processor architects will find the information useful since tolerance, avoidance, and elimination are generalizable constructsthat can serve as a basis for addressing other reliability challenges as well. Table of Contents: Introduction / Modeling Voltage Variation / Understanding the Characteristics of Voltage Variation / Traditional Solutions and Emerging Solution Forecast / Allowing and Tolerating Voltage Emergencies / Predicting and Avoiding Voltage Emergencies / Eliminiating Recurring Voltage Emergencies / Future Directions on Resiliency