A workshop for machine learning researchers This repository contains the materials used in the workshop and examples related to machine learning reproducibility.
Reproducibility is more than a technical checkbox: it is a habit of working well rooted in the principles of research integrity (Reliability, Honesty, Respect, Accountability, see this link). When you document your environment, structure your code, and track your decisions, you are not just helping others to build on your research: you are helping your future self, who will have forgotten why that choice was made and what that experiment was trying to prove.
The course follows a natural path that mirrors how good research actually happens:
- Plan — formalise your goal early. A model card written at the start ties together your environment, code, data, and intended outputs into a single checklist you fill in as you go.
- Work — make your own life easier. Version your environment, write modular code, and set up pipelines you can re-run without starting from scratch.
- Review — make it possible for others to verify and reproduce what you did, so the science can be checked.
- Publish — satisfy funder requirements and open your research outputs. Sharing a model responsibly means knowing what went into it.
See file 2026-05-20-schedule.md for the detailed schedule of the workshop.
See file 2025-11-11-schedule.md for the detailed schedule of the workshop.