How to Run the Examples in AI For Beginners: A Complete Setup Guide

Clone the Microsoft AI For Beginners repository, create the Conda environment from environment.yml, and run Python scripts with python or notebooks with jupyter notebook.

The AI For Beginners repository by Microsoft provides self-contained, beginner-friendly code examples that demonstrate core AI concepts without requiring the full curriculum notebooks. This guide walks you through running these examples step by step, from environment setup to executing your first machine learning model.

Prerequisites for Running AI For Beginners Examples

Before running any examples, ensure you have Python 3.8 or higher installed. The repository targets cross-platform compatibility across Windows, macOS, and Linux.

Two dependency management approaches are supported:

• Conda environment (recommended): Uses environment.yml to pin exact versions for reproducibility
• Minimal pip install: Lightweight setup for running only the Python scripts

The Conda approach automatically registers a Jupyter kernel, enabling direct notebook execution from the repository.

Creating the Conda Environment

The environment.yml file in the repository root defines an isolated environment named ai4beg with all required libraries: numpy, jupyter, tensorflow, pillow, and others.

Run these commands from the repository root:

git clone https://github.com/microsoft/AI-For-Beginners.git
cd AI-For-Beginners
conda env create --name ai4beg --file environment.yml
conda activate ai4beg

Alternative lightweight setup (Python scripts only):

pip install numpy jupyter pillow tensorflow

Running Python Script Examples

The examples/ directory contains four self-contained scripts, each runnable directly with the standard interpreter.

Hello AI World Demo

The 01-hello-ai-world.py script trains a minimal linear model (y = 2x) to introduce fundamental training concepts:

python examples/01-hello-ai-world.py

Expected output:

========================================
Welcome to Hello AI World!
========================================

📊 Creating training data...
Training examples: [(1, 2), (2, 4), (3, 6), (4, 8), (5, 10)]

🎓 Training started...
Epoch 20/100 - Average error: 0.4375 - Weight: 1.84
...
✅ Training complete! Final weight: 1.99

Each script includes a top-level main() guard (if __name__ == "__main__":) for direct execution.

Neural Network From Scratch

Run the multilayer perceptron implementation:

python examples/02-simple-neural-network.py

Text Sentiment Analysis

Execute the tiny sentiment model on sample sentences:

python examples/04-text-sentiment.py

Running Jupyter Notebook Examples

Some examples require interactive execution. The 03-image-classifier.ipynb notebook demonstrates image classification with a pre-trained CNN.

Launch the notebook server:

jupyter notebook examples/03-image-classifier.ipynb

Execution flow:

1. First cell: Import TensorFlow and NumPy
2. Second cell: Download and display a sample image
3. Third cell: Load MobileNet, run inference, and print top-5 predictions

The notebook interface renders results inline, making it ideal for visualizing model outputs.

Using VS Code Dev Containers or GitHub Codespaces

For zero-configuration execution, the repository includes .devcontainer/devcontainer.json. This configuration pre-installs the identical Conda environment in a Docker container.

Steps:

1. Open the repository in VS Code
2. Click "Reopen in Container" when prompted
3. The Jupyter server starts automatically

This approach eliminates local dependency management entirely.

Key Files for Running AI For Beginners Examples

File	Purpose
environment.yml	Conda specification with pinned Python and library versions
examples/01-hello-ai-world.py	Linear model trainer—your first "Hello AI" demo
examples/02-simple-neural-network.py	Feed-forward network implemented from scratch
examples/03-image-classifier.ipynb	Interactive image classification with pre-trained CNN
examples/04-text-sentiment.py	Sentiment analysis on short text inputs
.devcontainer/devcontainer.json	VS Code container definition for one-click setup

Summary

• Clone microsoft/AI-For-Beginners to your local machine
• Create environment using conda env create or minimal pip install
• Activate the ai4beg environment before running examples
• Run scripts with python examples/<script>.py
• Run notebooks with jupyter notebook <notebook>.ipynb
• Use dev containers for dependency-free execution in VS Code or GitHub Codespaces

All examples are deliberately isolated, heavily commented, and executable without the full curriculum—ready for hands-on AI exploration.

Frequently Asked Questions

Do I need to install the full curriculum to run the examples?

No. The examples in the examples/ directory are completely self-contained. You only need the files in that directory plus the dependencies specified in environment.yml. The full lesson notebooks are optional.

What if I don't want to use Conda?

You can run most examples with a minimal pip installation: pip install numpy for the basic scripts, or pip install numpy jupyter pillow tensorflow if you need notebook support. However, Conda is recommended for exact version matching.

Why does the image classifier use a pre-trained model instead of training from scratch?

The 03-image-classifier.ipynb notebook uses MobileNet, a pre-trained convolutional neural network, to demonstrate transfer learning—a practical technique where you leverage existing models rather than training computationally expensive networks from zero. This keeps execution fast and hardware requirements minimal for beginners.

Can I run these examples on Google Colab?

Yes, though you'll need to upload the notebook files manually. The 03-image-classifier.ipynb notebook will work in Colab with its pre-installed TensorFlow. For the .py scripts, upload them and run in Colab code cells or convert to notebooks using jupytext.

{
  "mcpServers": {
    "instagit": {
      "command": "npx",
      "args": ["-y", "instagit@latest"]
    }
  }
}