You have a research idea where you believe machine learning could help.
We offer fast-paced cooperation projects of up to 12 person-months to get you off the ground.
If you are a researcher in Tübingen, read on for details.
- 1️⃣ A research problem in the natural or social sciences, or the humanities, with
- a well-defined quantitative metric of performance. What kind of results would make the project a success?
- relevant references, including baselines that don't use ML techniques.
- (if you would like to learn about ML before engaging with us, we recommend this Introduction to Machine Learning series. Write to us for advice on other introductory material!)
- 2️⃣ A suitable dataset for machine learning,
- The bulk of the data must be available.
- Your dataset does not need to be big per se, but should be representative of the phenomenon.
- Knowledge of sampling biases and other dataset limitations is very helpful.
- If you don't know what kind of data you need to collect, we can set up a consultation session to advise.
- 3️⃣ One or more researchers with domain knowledge and enthusiasm to work collaboratively
- To facilitate skills transfer, at least one collaborator from your team should have some programming skills.
- 1️⃣ A quick evaluation of suitability: can ML really help you? If not, what is missing?
- We tell you in days if we have capacities for an evaluation.
- After an evaluation (<3 weeks) we move directly into the execution phase or make concrete recommendations. For example:
- if the nature of your project calls for it, we connect you to key partners from the ML community in Tübingen and translate your needs to them. You take it from there with collaboration arrangements, joint applications for funding, etc,
- or we tell you if more or other data is needed.
- 2️⃣ A close, iterative collaboration, "batteries included"
- We select, implement and run state-of-the-art ML algorithms on your dataset.
- All computations happen in our own ML Cloud — free of charge.
- We work iteratively and share progress via code repositories and collaboration platforms.
- Work starts shortly after evaluation; we don't prebook capacities.
- 3️⃣ Code, ideas, figures and writing for publications
- We stand for open science. Our deliverables, once released, are under a free content license and, by extension, are open access.
- You receive a (private) final report of activities including negative results ("what did not work"), guidance for further ML work in your group, and suggestions for further work (that can be used to apply for follow-ups).