Application Development Workshop with Machine Learning

Comprehensive practice questions covering Google Colab, Machine Learning fundamentals, TensorFlow/Keras, linear regression, Teachable Machine, and edge deployment for software and hardware integration.

What is Google Colab (Google Colaboratory)?

A free cloud-based platform provided by Google that allows users to write and execute Python code in a browser, based on Jupyter Notebooks and integrated with Google Drive.

What are the primary characteristics of Google Colab?

It is cloud-based with no installation required, compatible with Jupyter notebooks ($.ipynb$), and provides free access to GPUs and TPUs for Machine Learning.

How can personal libraries be installed in Google Colab?

By using the command $!pip \, install$.

What are the advantages of working in collaborative cloud environments like Colab?

Access from anywhere with internet, real-time collaboration, easy sharing like Google documents, automatic security/backup in Drive, and no local installation needed.

What are the two types of cells used in Google Colab?

Code cells (to write and execute Python) and Text cells (Markdown for documentation, titles, lists, and equations).

What are the functions of NumPy, Pandas, and Matplotlib in Machine Learning?

NumPy for efficient numerical calculations with matrices; Pandas for data manipulation using DataFrames; Matplotlib for data visualization through various charts.

What are some common Pandas operations for data manipulation?

Column selection, row filtering, value modification, grouping ($df.groupby()$), handling null values ($df.dropna()$, $df.fillna()$), and statistical summaries ($df.describe()$).

How do CPU, GPU, and TPU differ in Google Colab?

CPU is for general tasks; GPU is designed for parallel operations and matrix calculations (ideal for ML/Deep Learning); TPU is a specialized processor created by Google exclusively for TensorFlow operations.

How can a user change the hardware accelerator in Google Colab?

Go to the 'Runtime' (Entorno de ejecución) menu, click 'Change runtime type', and select None (CPU), GPU, or TPU under 'Hardware accelerator'.

What code can be used to monitor RAM and Disk usage in Google Colab?

Using $psutil.virtual\_memory()$ for RAM and $shutil.disk\_usage('/')$ for Disk.

What is the difference between Supervised and Unsupervised learning?

Supervised learning uses labeled data where the model knows the expected answer; Unsupervised learning explores data without labels to find hidden patterns or groupings.

What are the 6 steps in the Machine Learning training workflow?

1. Data collection, 2. Preprocessing (cleaning/normalizing), 3. Dataset division (train/test), 4. Model definition, 5. Training, 6. Evaluation and improvement.

What is the difference between Sequential and Functional models in Keras?

Sequential models are linear (layer by layer) for simple problems; Functional models are flexible for complex architectures with multiple inputs or outputs.

What is the general formula for Simple Linear Regression?

$y = mx + b$, where $y$ is the dependent variable, $x$ is the independent variable, $m$ is the slope, and $b$ is the intercept.

What are the common metrics used to evaluate model performance?

MSE (Mean Squared Error), RMSE (Root Mean Squared Error), and $R^2$ Score (Coefficient of Determination).

How is $R^2$ Score interpreted?

It indicates how well the model explains variability, ranging from 0 to 1; a value closer to 1 signifies a good model.

What is Teachable Machine?

A free browser-based Google tool that allows users to create Machine Learning models (Image, Audio, Poses) without needing to program.

What three types of models are supported by Teachable Machine?

Image (recognizes objects), Audio (recognizes words/sounds), and Poses (detects body positions).

In which formats can models be exported for different environments?

TensorFlow.js ($.json$ and binary for web), Keras HDF5 ($.h5$ for Python), and TensorFlow Lite (optimized for mobile/embedded devices).

What are the characteristics of TensorFlow.js?

It uses JavaScript, runs directly in the browser (client-side), and is compatible with models trained in Teachable Machine or TensorFlow.

What is OpenCV used for in Python?

It is an open-source library for image processing and computer vision tasks like edge detection ($cv2.Canny$) and face detection ($cv2.CascadeClassifier$).

How does TensorFlow Lite (TFLite) improve model deployment on small devices?

It provides model quantization to reduce size and increase inference speed with minimal precision loss, making it ideal for mobile and IoT devices.

What is TinyML?

A branch of AI focused on running Machine Learning models on devices with very low resources, such as microcontrollers, featuring low power consumption and local execution.

What is the workflow for deploying AI on an Arduino Nano 33 BLE Sense?

Collect data from sensors, train the model (Colab or Edge Impulse), convert it to TFLite, then convert it to C++ to integrate with the Arduino code.

What is Edge Computing in the context of Raspberry Pi?

Processing data directly on the device (the 'edge') rather than depending on the cloud, used for real-time tasks like facial detection.