NEW
Font size
WorksheetsCOE64-351_Quiz 3
Total questions: 20
Worksheet time: 10mins
What is the primary objective of Machine Learning?
To explicitly program computers for every possible scenario.
To enable computer systems to learn from data, identify patterns, and make decisions without explicit programming.
To increase the processing speed of computer hardware.
To collect and store large amounts of data from IoT sensors.
In a predictive maintenance project for industrial machines, what would the 'Label' typically represent?
The real-time temperature and vibration sensor readings.
The total number of hours the machine has been in operation.
A binary outcome, such as 'Normal' or 'Faulty'.
The algorithm used to train the model, such as a Decision Tree.
A smart home system learns to adjust the thermostat based on receiving positive feedback (rewards) for saving energy and negative feedback (penalties) for wasting it. Which type of Machine Learning is this?
Supervised Learning
Unsupervised Learning
Reinforcement Learning
Regression Learning
What does the term TinyML represent?
A simplified version of Machine Learning that only works on mobile phones.
The intersection of Machine Learning algorithms and low-power, embedded IoT devices.
A cloud computing service for training very small machine learning models.
A type of hardware specifically designed for AI.
Which of the following is a primary advantage of using TensorFlow Lite for on-device machine learning?
It requires a constant, high-speed internet connection.
It can only be used with the Python programming language.
It is optimized to address constraints like latency, privacy, and power consumption.
It automatically collects and labels new training data from the device.
A developer trains a model to detect anomalies in factory machinery. It achieves 99% accuracy on the training data, but only 65% accuracy on new, unseen testing data. What is the most likely problem?
Underfitting
Overfitting
Lack of Features
Incorrect Algorithm Choice
In the TinyML workflow, what is the purpose of techniques like 'pruning' and 'quantization'?
To collect more diverse training data from sensors.
To select the best machine learning algorithm for a task.
To increase the complexity and accuracy of the cloud-based model.
To compress and optimize a trained model to run efficiently on a resource-constrained microcontroller.
Which machine learning task would be used to predict the exact energy consumption (a continuous value in kWh) of a smart building?
Classification
Clustering
Regression
Reinforcement Learning
Why is 'Real-Time Decision Making' a key reason for using ML in IoT applications like smart city traffic management?
Because it allows the city to store historical traffic data for later analysis.
Because ML models can adapt to new traffic patterns over several months.
Because ML can process sensor data instantly to make rapid adjustments, like changing traffic signals to ease congestion.
Because it helps in personalizing routes for individual drivers.
Which of these is NOT a primary characteristic of a TinyML application?
Minimal power consumption (<1mW)
Dependence on a constant cloud connection
Enhanced data privacy
Accelerated decision-making
Which type of algorithm would be most suitable for an IoT application that needs to group smart home devices based on their energy consumption patterns, without any pre-existing labels?
Classification
Regression
Reinforcement Learning
Clustering
What is the purpose of a 'Testing Dataset' in the machine learning workflow?
To train the machine learning model on a large set of labeled examples.
To adjust the model's hyperparameters for optimal performance.
To provide a final, unbiased evaluation of the model's performance on new, unseen data.
To group similar data points together without explicit guidance.
The TinyML architecture shows that after a model is trained, it undergoes a 'Post-Training Model Processing' stage. Which of the following happens during this stage?
The model is deployed on the microcontroller.
Raw signals are collected from the sensors.
The model is converted, compressed, and quantized.
The model's accuracy and F1 score are calculated.
For a real-time, battery-powered medical wearable that detects heart anomalies, why is TinyML a fundamentally better architectural choice than a cloud-based ML approach?
Because cloud models are always less accurate than TinyML models.
Because cloud platforms like AWS and Azure are not secure.
Because it guarantees low latency, high privacy, and offline functionality, which are critical for this application.
Because it eliminates the need for any data storage.
Based on the table 'ML algorithms used in internet of things (IoT) applications', which ML algorithm demonstrates the widest applicability across BOTH 'Classification' and 'Prediction' tasks?
K-Means
Recurrent Neural Network (RNN)
Support Vector Machine (SVM)
Q-Learning
What is the key difference between 'Features' and 'Labels' in supervised machine learning?
Features are numerical data, while Labels are categorical data.
Features are the input variables for the model, while the Label is the output variable the model learns to predict.
Features are used in the training phase, while Labels are only used in the testing phase.
There is no difference; the terms are used interchangeably.
Which of these steps comes LAST in the TinyML workflow before the device can make an inference?
Building and Training the ML Model
Pruning and Knowledge Distillation
Deploying the Model on the Microcontroller (MCU)
Feature Extraction from Sensor Data
A smart irrigation system is designed to either 'Water' or 'Not Water' based on sensor readings. What kind of ML algorithm is most appropriate for this task?
Clustering
Classification
Regression
Anomaly Detection
According to the workflow diagram, the ML process is iterative. What action directly follows 'Get Feedback'?
Train Model
Predict
Generate New Data
Perform an action
If you are designing an IoT system to sort fruit on a conveyor belt into 'ripe', 'unripe', and 'rotten' categories based on camera images, which combination of ML branch and task is most accurate?
Unsupervised Learning - Clustering
Supervised Learning - Regression
Reinforcement Learning - Decision Making
Supervised Learning - Classification
