Hands-on Workshop · 8 Jupyter Notebooks
Bienvenue ! This README gathers everything you need before arriving at the
90-minute workshop. Follow the steps for your operating system and
you’ll be ready to run all notebooks offline — the data are already bundled
inside the repo (./data/).
| Task | macOS 13 + | Windows 10/11 | Ubuntu 22.04 |
|---|---|---|---|
| Install Anaconda | .pkg |
.exe |
.sh |
Create env sleep‑ai & deps |
Terminal.app | Anaconda Prompt | Terminal |
| Launch JupyterLab | conda activate … && jupyter lab |
same | same |
| Clone repo | git clone … |
Git Bash / Desktop | same |
Total prep time ≈ 15 min.
- Load & pre‑process EEG/ECG/EDA
- Engineer classical features & visualise
- Train ML models (LogReg, SVM, RF)
- Compare to a tiny raw‑signal CNN (TF & PyTorch)
| # | Notebook | Core concept | Time |
|---|---|---|---|
| 01 | 01_Intro.ipynb | PSG loading & band‑power | 12 min |
| 02 | 02_Sleep_Stage_Logreg.ipynb | Logistic Regression | 15 min |
| 03 | 03_KMeans_HRV.ipynb | HRV K‑Means | 12 min |
| 04 | 04_SVM_EDA_Arousal.ipynb | EDA SVM | 15 min |
| 05 | 05_PCA_UMAP_Fusion.ipynb | PCA + UMAP | 12 min |
| 06 | 06_RandomForest_SE.ipynb | Sleep Efficiency RF | 12 min |
| 07a | 07a_Tiny_CNN_TF.ipynb | CNN (TensorFlow) | 15 min |
| 07b | 07b_Tiny_CNN_PT.ipynb | CNN (PyTorch) | 15 min |
Choose the installer for your OS:
- macOS Apple Silicon 🤍
Anaconda3‑2025‑MacOS‑arm64.pkg - macOS Intel
…‑x86_64.pkg - Windows
Anaconda3‑2025‑Windows‑x86_64.exe - Ubuntu
Anaconda3‑2025‑Linux‑x86_64.sh
conda create -n sleep-ai python=3.11 -y
conda activate sleep-ai
conda install numpy pandas scipy scikit-learn matplotlib seaborn -y
pip install mne neurokit2 umap-learn imbalanced-learn
# deep‑learning
pip install tensorflow==2.16 # notebook 07a
pip install torch==2.2 torchinfo # notebook 07bApple Silicon GPU
pip install --pre torch --extra-index-url https://download.pytorch.org/whl/nightly/cpu
export PYTORCH_ENABLE_MPS_FALLBACK=1conda install jupyterlab notebook -y
jupyter lab # opens http://localhost:8888workshops/
├─ data/ # contains ST7011J0 EDF files
├─ notebooks/ # eight .ipynb notebooks
├─ outputs/ # CSVs generated during the lab
└─ README.md
Clone the repo:
git clone https://github.com/ppsp-team/workshops
cd workshopsNo extra download — the EDF files are already in ./data/.
| Tip | macOS M‑series | Windows/Linux (NVIDIA) |
|---|---|---|
| TensorFlow GPU | pip install tensorflow‑metal |
install CUDA 12.4 |
| PyTorch device | device="mps" |
device="cuda" |
| DataLoader workers | num_workers=4 |
num_workers=4‑8 |
| Mixed precision | autocast("mps") |
autocast() |
| Symptom | Fix |
|---|---|
ModuleNotFoundError |
pip install <package> |
| Kernel not in Jupyter | python -m ipykernel install --user --name sleep-ai --display-name "Python (sleep-ai)" |
| Slow PyTorch on Mac | use mps, workers, avoid per‑batch .to(device) |
| Matplotlib backend error (Win) | pip install pyqt5 or %matplotlib inline |
Questions? Open an issue or email remy.ramadour.hsj @ ssss.gouv.qc.ca.
See you at the Summer‑School — happy coding!