Skip to content
book.derekmolloy.ie

Search is only available in production builds. Try building and previewing the site to test it out locally.

📝 Knowledge Test

Chapter 2 Knowledge Test

Section titled “Chapter 2 Knowledge Test”
/
Knowledge Test Score Board
—· 0%

Work through the questions below — your score updates as you go.

Keep going · grade bands mapped to DCU's honours scale

target15:00

Only your first attempt at each question counts toward this score. Reload the page to reset.

This section contains 20 interactive knowledge checks designed to test your understanding of the materials covered in Chapter 2. The questions span foundational edge computing concepts, state-of-the-art hardware, programming best practices, and distributed systems orchestration. The section includes some inheritance hierarchy questions to further help you digest the materials from Chapter 1.

Please remember that many of the quizzes have multiple correct answers, and you must select all applicable options to succeed. Take your time to carefully evaluate each option.

Good luck!

Derek.

Part 1: Core Concepts and Edge Constraints

Section titled “Part 1: Core Concepts and Edge Constraints”
Q1
Quiz
Select 0/3

Which of the following represent the primary drivers for transitioning computational tasks from the centralised cloud to the network edge?

Enhanced data privacy and security by processing sensitive biometric or industrial information locally, preventing exposure during transmission.
Unlimited computational scalability by leveraging local area networks to permanently store and archive all historical raw data across the edge fleet.
Decreased latency for real-time decision-making, ensuring that immediate actions can be taken without unpredictable network delays.
Reduced bandwidth consumption by filtering and analysing data at the source, transmitting only essential insights or model updates to the cloud.
Q2
Concept Match

Match the Edge AI Term to its Definition

Drag each definition into its matching concept slot, then click Submit. Tap × to return a placed card to the pool.

TinyML
drag a definition here…
Model Drift
drag a definition here…
Catastrophic Forgetting
drag a definition here…
Offline-First Design
drag a definition here…

Definition Pool

Enables machine learning inference directly on low-power microcontrollers, often consuming less than 1 milliwatt of power.
A scenario where a highly constrained model loses previously learned information while attempting to learn new data.
Engineering applications to gracefully handle extended periods of disconnection using local data caching and processing.
The gradual decline in an AI model's accuracy because the real-world data distribution changes over time.
Q3
Quiz
Select 0/2

When distinguishing between the tiers of a modern edge computing architecture, which statements accurately describe their roles?

Edge servers are ultra-low-power microcontrollers running on coin-cell batteries, deployed directly onto moving industrial machinery.
Edge servers act as intermediary nodes that aggregate data from multiple endpoints to handle heavier computational AI workloads locally.
Endpoint devices serve as centralised data lakes that orchestrate long-term machine learning model training across the entire global network.
Endpoint devices are highly constrained sensors designed for ultra-low-power inferencing directly at the exact source of data generation.
Q4
Abstract Class Hierarchy

Build the Edge Architecture Hierarchy

Drag class names from the pool into the slots so each class sits below its parent in the inheritance tree. Abstract classes are shown in italics with an amber accent — slots expecting an abstract class are pre-marked with «abstract». Click Submit when you're done.

Inheritance Tree (base class on top)
«abstract»
«abstract»
·····
·····
«abstract»
·····

Available Classes

PowerGrid
DataPacket
NetworkCable
SensorReading
SystemNode
EdgeNode
AnalyticsServer
EndpointDevice
CloudNode
EdgeServer
Q5
Quiz
Select 0/1

Why is the volatility of SRAM a significant challenge for TinyML devices?

It generates excessive thermal heat, requiring microcontrollers to use active liquid cooling systems in industrial environments.
It permanently corrupts the pre-trained neural network weights stored in the flash memory during every read operation.
It requires the device to maintain a constant, high-bandwidth 5G connection to stream memory states to the cloud.
It means that any on-device training progress or dynamically learned state is completely lost upon a power-off or hardware reset.

Part 2: State-of-the-Art Hardware

Section titled “Part 2: State-of-the-Art Hardware”
Q6
Concept Match

Match the Hardware Architecture to its Strength

Drag each definition into its matching concept slot, then click Submit. Tap × to return a placed card to the pool.

GPU
drag a definition here…
NPU
drag a definition here…
FPGA
drag a definition here…
ASIC
drag a definition here…

Definition Pool

A dedicated co-processor specifically designed to accelerate matrix multiplication and tensor operations with high efficiency.
A custom-made silicon chip offering a superior price-to-performance ratio and exceptional power efficiency for a fixed, specific workload.
A reprogrammable logic fabric offering unparalleled in-field reconfigurability and deterministic low-latency performance.
A parallel processing powerhouse highly effective for deep learning, computer vision, and generative AI workloads.
Q7
Quiz
Select 0/3

Which of the following characteristics accurately describe Neuromorphic Computing (such as Intel's Loihi 2)?

It operates asynchronously, mimicking the biological brain, which leads to unparalleled energy efficiency and low response latencies.
It relies entirely on synchronous, high-frequency clocking to guarantee the deterministic execution of massive, parallel deep neural networks.
It employs a highly efficient event-driven architecture that only activates processing when specific changes or spikes occur in the input data.
It integrates memory and processing directly into a single unified architecture, minimising the energy loss associated with constant data movement.
Q8
Abstract Class Hierarchy

Build the Processor Hierarchy

Drag class names from the pool into the slots so each class sits below its parent in the inheritance tree. Abstract classes are shown in italics with an amber accent — slots expecting an abstract class are pre-marked with «abstract». Click Submit when you're done.

Inheritance Tree (base class on top)
«abstract»
«abstract»
·····
«abstract»
·····
·····

Available Classes

CoolingFan
SpecialisedAccelerator
Motherboard
GeneralPurposeCPU
NeuralProcessingUnit
Processor
ClockSignal
NeuromorphicChip
Microcontroller
SiliconWafer
Q9
Quiz
Select 0/2

Regarding the ESP32 family of microcontrollers, which statements are true?

The ESP32 Classic relies exclusively on a powerful GPU architecture to render complex 3D graphics for embedded touch displays.
The ESP32-S3 features advanced vector processing extensions that significantly accelerate the integer arithmetic used in AI inference.
The ESP32-C3 utilises an open-source RISC-V core and serves as a highly cost-effective, single-core alternative for simple IoT nodes.
The ESP32-H2 is designed specifically to run full desktop operating systems like Windows 11 due to its high base clock speeds.
Q10
Concept Match

Match the AI Model Optimisation Technique

Drag each definition into its matching concept slot, then click Submit. Tap × to return a placed card to the pool.

Quantisation
drag a definition here…
Pruning
drag a definition here…
Knowledge Distillation
drag a definition here…
Fine-tuning
drag a definition here…

Definition Pool

A critical recovery step applied after compression to retrain the model and recover any lost predictive accuracy.
Systematically identifying and removing unnecessary neurons or connections to create a sparser, faster network.
Training a compact 'student' model to emulate the output behaviour of a massive, complex 'teacher' model.
Reducing the mathematical precision of weights and activations (e.g., from 32-bit float to 8-bit integer) to save memory.

Part 3: Programming Best Practices

Section titled “Part 3: Programming Best Practices”
Q11
Quiz
Select 0/2

How does the adoption of 'DevSecOps' and the 'shift left' approach improve embedded software development?

DevSecOps integrates development, security, and operations, promoting automation and collaboration across edge and cloud environments.
The 'shift left' approach ensures that software testing, validation, and security considerations are incorporated much earlier in the development lifecycle.
The Waterfall methodology is strictly enforced to ensure that all system requirements are permanently frozen before any coding or testing begins.
Security penetration testing is exclusively delayed until the final deployment phase to avoid interrupting the initial software development sprints.
Q12
Class Hierarchy

Build the Operating Environment Hierarchy

Drag class names from the pool into the slots so each class sits below its parent in the inheritance tree, then click Submit.

Inheritance Tree (base class on top)
·····
·····
·····
·····
·····

Available Classes

ManagedOS
RTOS
BinaryExecutable
JTAGDebugger
Compiler
EmbeddedLinux
OperatingEnvironment
BareMetal
FileSystem
Q13
Quiz
Select 0/2

Which of the following are recommended best practices for Memory and Power Management in edge systems?

Developers should rely entirely on dynamic memory allocation (malloc/new) throughout runtime to ensure maximum memory flexibility.
Event-driven processing ensures the CPU runs continuously at maximum frequency to guarantee that no rapid sensor events are missed.
Dynamic Voltage and Frequency Scaling (DVFS) adjusts the processor's voltage and clock speed dynamically based on the current workload.
Using pre-allocated memory pools with robust failsafes provides operational flexibility while maintaining strict determinism and preventing crashes.
Q14
Concept Match

Match the Embedded Engineering Tool

Drag each definition into its matching concept slot, then click Submit. Tap × to return a placed card to the pool.

RTOS
drag a definition here…
DMA
drag a definition here…
HIL Simulation
drag a definition here…
JTAG
drag a definition here…

Definition Pool

A standard hardware interface used for low-level in-circuit debugging, boundary scanning, and memory inspection.
An operating system that guarantees deterministic, predictable timing for critical task execution.
A hardware feature allowing peripherals to transfer data directly to memory without continuous CPU intervention.
Validating firmware logic by running it against a simulated or physical hardware environment in real-time.
Q15
Quiz
Select 0/2

Which strategies align with the principle of 'Security by Design' in embedded edge systems?

Enforcing the principle of 'keeping the edges dumb' by limiting sensitive data storage and processing it in highly secure server environments.
Transmitting all biometric sensor data in plain text to significantly reduce the computational overhead and power draw of encryption algorithms.
Disabling all forms of over-the-air (OTA) updates permanently to ensure that malicious actors cannot remotely modify the firmware after deployment.
Implementing secure boot mechanisms to ensure that only cryptographically signed and trusted firmware can be executed by the hardware.

Part 4: Distributed Programming and Orchestration

Section titled “Part 4: Distributed Programming and Orchestration”
Q16
Quiz
Select 0/2

What makes Federated Learning (FL) a transformative approach for distributed AI on edge devices?

It involves iterative cycles where local model updates (weight and parameter changes) are sent to a central server for aggregation into an improved global model.
It mandates that all edge devices continuously upload their complete, raw datasets to a central cloud server to guarantee absolute model accuracy.
It is exclusively used for wired industrial automation protocols and is legally prohibited from being applied to privacy-sensitive healthcare applications.
It enables collaborative machine learning model training without compromising local data privacy by keeping sensitive raw data strictly on-device.
Q17
Abstract Class Hierarchy

Build the Communication Protocol Hierarchy

Drag class names from the pool into the slots so each class sits below its parent in the inheritance tree. Abstract classes are shown in italics with an amber accent — slots expecting an abstract class are pre-marked with «abstract». Click Submit when you're done.

Inheritance Tree (base class on top)
«abstract»
«abstract»
·····
·····
«abstract»
·····

Available Classes

EthernetCable
Thread
NetworkProtocol
MQTT
BaudRate
FrequencyBand
MessagingProtocol
WirelessStandard
Antenna
Matter
Q18
Concept Match

Match the Distributed Orchestration Concept

Drag each definition into its matching concept slot, then click Submit. Tap × to return a placed card to the pool.

MQTT
drag a definition here…
Matter
drag a definition here…
K3s
drag a definition here…
WebAssembly (Wasm)
drag a definition here…

Definition Pool

A secure, platform-independent sandbox offering near-native execution speeds with a significantly smaller footprint than traditional containers.
A unifying application-layer standard that ensures seamless interoperability between IoT devices from different manufacturers.
A lightweight publish/subscribe messaging protocol designed for reliable communication in low-bandwidth IoT environments.
A highly simplified, lightweight Kubernetes distribution explicitly optimised for managing containers on constrained edge devices.
Q19
Quiz
Select 0/2

In the context of workload orchestration, what are the distinct advantages of K3s and WebAssembly (Wasm) at the edge?

K3s provides a fully compliant Kubernetes cluster distributed as a single binary, radically reducing the memory footprint and CPU overhead required for orchestration.
Service Mesh technologies are used to physically wire K3s and Wasm nodes together using proprietary industrial Ethernet cables for maximum throughput.
WebAssembly (Wasm) offers a highly secure sandbox for executing 'serverless' edge functions, making it ideal for ultra-constrained nodes where even K3s is too heavy.
Both K3s and WebAssembly require massive, liquid-cooled data centre infrastructure, meaning they are strictly deployed in the central cloud, never the edge.
Q20
Abstract Class Hierarchy

Build the Optimisation Strategy Hierarchy

Drag class names from the pool into the slots so each class sits below its parent in the inheritance tree. Abstract classes are shown in italics with an amber accent — slots expecting an abstract class are pre-marked with «abstract». Click Submit when you're done.

Inheritance Tree (base class on top)
«abstract»
«abstract»
·····
·····
«abstract»
·····

Available Classes

Pruning
Overclocking
Quantisation
SRAM
DeepLearning
OptimisationStrategy
KnowledgeDistillation
LiquidCooling
SizeReduction
ArchitectureChange

Congratulations on completing the Chapter 2 Knowledge Test! Review any questions you missed to ensure a solid foundation in hardware architectures and edge constraints before diving into the programming tutorials.

© 2026 Derek Molloy, Dublin City University. All rights reserved.