ML Reproducibility Workshop for AI-DOC (slides)

I organized and taught a workshop on “Reproducibility and Open Science Practices in Machine Learning” for the AI-DOC doctoral program at Aalto University.

The workshop addressed the fundamental question: what is reproducibility and why do we need to care about it in machine learning research?

Topics covered:

• Environment setup with containers and package management
• Reproducible Jupyter notebooks
• Experiment tracking with MLflow
• Modular training pipelines with PyTorch Lightning
• Model sharing via Hugging Face
• Distributed training and multi-GPU optimization
• Overfitting prevention

Collaborators: Simo Tuomisto, Hossein Firooz, Luca Ferranti

Resources:

• Slides at Zenodo (CC BY 4.0)
• Hands-on exercises at GitHub