The Seven Layers of AI, Layer 2: Machine Learning

This notebook accompanies Chapter 3, Layer 2: Machine Learning of the study guide The Seven Layers of AI. It contains every example program from the chapter in the order it appears, so you can run, change and experiment with the code as you read.

What is in the notebook

For each program you will find:

1. The explanation from the guide that introduces the program.
2. The code, with a header explaining its purpose, how it works and what to look for, plus comments throughout.
3. The output from a test run, already saved in the notebook, so you can read it before running anything.
4. The output shown in the guide, for comparison. The two should match, apart from timings.
5. What the output shows, a short interpretation.
6. Investigate tasks, which suggest changes to make and questions to answer.

Most programs use scikit-learn's built-in datasets, so nothing needs downloading. l2_pipeline.py saves a trained model with joblib into the folder the notebook runs from.

Programs

#	File	Topic	Status
1	l2_workflow.py	the machine learning workflow, and two ways to fool yourself	runs in the notebook
2	l2_linear_regression_worked.py	linear regression, calculated four ways	runs in the notebook
3	l2_logistic.py	classifying tumours	runs in the notebook
4	l2_trees.py	decision trees, random forests and gradient boosting	runs in the notebook
5	l2_knn_svm.py	k-NN from scratch and SVM kernels	runs in the notebook
6	l2_naive_bayes.py	a spam filter	runs in the notebook
7	l2_unsupervised.py	k-means from scratch and PCA	runs in the notebook
8	l2_overfitting.py	overfitting and cross-validation	runs in the notebook
9	l2_evaluation.py	the accuracy trap	runs in the notebook
10	l2_pipeline.py	an end-to-end churn model	runs in the notebook
11	l2_qlearning.py	Q-learning in a grid world	runs in the notebook
12	l2_fairness.py	bias through a proxy	runs in the notebook

Requirements

• Python 3.10 or later
• Jupyter (JupyterLab, Jupyter Notebook, or VS Code with the Jupyter extension)
• Packages: numpy pandas scikit-learn matplotlib joblib

The notebook takes about 2 minutes to run from top to bottom on a typical laptop.

Setting up

These steps create a separate Python environment so the packages do not interfere with anything else on your computer. Run them in a terminal, from the folder that contains this README.

Windows (PowerShell)

python -m venv .venv
.venv\Scripts\Activate.ps1
pip install jupyterlab numpy pandas scikit-learn matplotlib joblib
jupyter lab

macOS or Linux

python3 -m venv .venv
source .venv/bin/activate
pip install jupyterlab numpy pandas scikit-learn matplotlib joblib
jupyter lab

JupyterLab opens in your browser. Double-click Layer2_Machine_Learning.ipynb to open it.

VS Code: open this folder, open Layer2_Machine_Learning.ipynb, and choose the .venv environment when VS Code asks you to select a kernel. Install the packages into that environment first, as above.

Google Colab: upload Layer2_Machine_Learning.ipynb at https://colab.research.google.com. Most packages are already installed; if one is missing, remove the # from the %pip install line in the setup cell and run it.

Using the notebook

1. Run the setup cell first. It is the first code cell. Uncomment the %pip install line if you have not installed the packages.
2. Run cells from top to bottom with Shift+Enter, or use Run All. Each program is self-contained, so after the setup cell you can also jump to any program and run just that one.
3. Read the saved output before running. It shows what to expect. When you run a cell, the saved output is replaced with yours.
4. Compare with the guide. The programs use fixed random seeds, so your numbers should match the guide exactly. Only timings, such as the seconds reported while training, vary between computers.
5. Work through the Investigate tasks. Edit the code directly in the cell and run it again. If you want to get back to the original, the code is also printed in the guide.
6. Restart when things get confusing. All programs share one Python session, so a variable changed in one cell can affect a later one. Kernel > Restart Kernel and Clear Outputs gives you a clean start.

Troubleshooting

Problem	What to do
ModuleNotFoundError: No module named ...	The package is not installed in the environment the notebook is using. Run %pip install <package> in a cell, then restart the kernel.
A cell has been running for a long time	Training programs can take a minute or more. Watch the [*] marker beside the cell; use Kernel > Interrupt to stop it.
Numbers differ slightly from the guide	Check you have not changed a seed or parameter. Very different package versions can also change results slightly; the notebooks were tested with NumPy 2.4, scikit-learn 1.8 and pandas 3.0.
FileNotFoundError for a data file	Run the cell that creates the file first (for example, the %%writefile cell), and keep the notebook in the folder it runs from.

Using this with learners

• Ask learners to predict the output before running a program, then explain any differences.
• Most Investigate tasks can be completed in 10 to 20 minutes and work well as paired activities.
• Encourage learners to break the code deliberately (change a learning rate, remove a line) and explain what happens. It is often the fastest way to understand why each line is there.

Source

Programs and explanations are from The Seven Layers of AI study guide, Chapter 3. Citations in the code and text refer to the guide's reference list.

License

Copyright © 2026 Trevor Smith. All rights reserved. Released under a Personal Use License.

In plain terms:

• ✅ You can download it from this repository and use it for free, for personal use, study or at work (including teaching).
• ✅ You can change it for your own private use.
• ❌ You can't share, copy, upload or distribute it, modified or not. To tell someone about it, send them a link to this repository.
• ❌ You can't sell it, or charge for a product or service based on it.
• ❌ You can't remove the copyright notice or present it as your own work.

The full LICENSE text is the legally binding version; this summary is just a guide. This is proprietary material, not open source. It replaces the earlier MIT and CC BY-NC 4.0 licences.