This repository contains a set of tutorials which are the introduction for the topics covered in the subject Machine Learning II of the Master in Artificial Intelligence taught at the three universities of Galicia: University of A Coruña (UDC), University of Santiago de Compostela (USC), and University of Vigo (UVigo).

• Teaching Staff
• Development Environment
  • Quick guide: package and environment management with uv
  • Install Required Packages
• Notebooks
• Jupyter
• Troubleshooting
• Notes

Teaching Staff:

• David Mera Pérez (coordinator, USC)
• Enrique Fernández Blanco (UDC)
• David Olivieri Cecchi (UVigo)

Development Environment

The practical sessions will be developed via Notebooks. To run them you will need Python (>3.8), a Notebook server (e.g., Jupyter), and all the necessary packages. There are different configuration options depending on your preferences:

1. Google Colab: Run the Notebooks online and install extra libraries using commands such as !pip install xxx. The free tier has some limitations, mainly related to computing resources.

2. Python + pip + virtual environments: Install Python (>3.8) from the official website. Dependencies can be installed using pip inside a virtual environment.

3. uv: A fast Python package and environment manager that replaces conda for this course. It creates standard virtual environments and uses pip-compatible workflows.

---

Quick guide: package and environment management with uv

This project uses uv as an alternative to conda for managing Python environments and dependencies.
uv creates standard Python virtual environments, so the usual pip workflows apply.

Create the virtual environment

From the root of the repository:

uv venv

This creates a standard virtual environment in .venv/.

---

Activate the environment

uv does not provide its own activate command.
Activation is done using the standard Python venv scripts.

macOS / Linux

source .venv/bin/activate

Windows (PowerShell)

.\.venv\Scripts\Activate.ps1

---

Install all project dependencies

With the corresponding environment activated, all dependencies are defined in requirements.txt and can be installed.

pip install -r requirements.txt

Since the environment was created with uv, this pip command installs packages inside the project virtual environment.

---

Install a single package

If the environment is activated:

pip install package_name

Example:

pip install river

---

Install packages without activating the environment (alternative)

You can also use uv directly:

uv pip install package_name

Examples:

uv pip install river
uv pip install "flwr[simulation]"
uv pip install -r requirements.txt

This automatically installs packages into the project virtual environment (.venv).

---

Run commands inside the environment (without activation)

Instead of activating the environment, you may use:

uv run command

Example:

uv run jupyter notebook

This ensures the command runs using the project virtual environment.

---

Summary

Task	Recommended command
Create environment	uv venv
Activate environment	source .venv/bin/activate
Install dependencies	pip install -r requirements.txt
Install one package	pip install package_name
Install without activation	uv pip install package_name
Run command without activation	uv run command

Install Required Packages

The table below summarizes the main libraries used in the course. You do not need to install them one-by-one (they are already included in requirements.txt), but this overview can help you understand what each unit relies on.

Package	Used for	Notes
jupyter / ipykernel	Running notebooks locally	Optional kernel registration supported
numpy, pandas	Data manipulation	Core scientific stack
matplotlib	Plotting	Figures in notebooks
scikit-learn==1.7.2	Classical ML + utilities	Pinned for River compatibility
river	Online learning + concept drift	Early units (Online ML)
flwr[simulation]	Federated learning simulations	Simulation extras required
tensorflow	Federated learning labs	Model training backend
ray[default]	Flower simulation backend	Required by Flower simulation mode
python-graphviz	Visualization	Some notebooks use Graphviz
rich	Pretty console output	Used in some examples

---

Notebooks

To obtain the notebooks for laboratory practices, you can either download the ZIP file from GitHub or clone the repository using Git:

git clone git@github.com:ennanco/MIA_ML2.git

Important: The examples located within the initial three working units (online ML + Concept Drift) have been specifically tailored for compatibility with River 0.22.

---

Jupyter

With the virtual environment activated, run:

jupyter notebook

Then open your browser at http://localhost:8888/.

The security token will be shown in the terminal.

Optional: Install a dedicated kernel

python -m ipykernel install --user --name mia-ml2 --display-name "Python (MIA-ML2)"

This allows selecting the correct environment from within Jupyter.

---

Troubleshooting

• River’s wrapper module, designed for integration with libraries like scikit-learn, currently has compatibility issues. Scikit-Learn versions ≥ 1.8.0 exhibit integration problems with River, resulting in the following error in Unit02:

'super' object has no attribute '__sklearn_tags__'

Until River is updated, the recommended workaround is to downgrade scikit-learn:

pip uninstall scikit-learn
pip install "scikit-learn==1.7.2"

---

Notes

• Always ensure you are using the pip from the active virtual environment.
• You may also use Google Colab if you prefer not to install anything locally.