The thread: five steps, one way of working

Every piece of research starts with the same question: what do we already know? Depending on the answer, compute, complete a model or learn from data; then make the tool usable and demonstrate that it works. Follow the thread down: each step leads to its stories, each on its own page.

Step 1 of 5 · Is the link between causes and effects known?

Compute directly

Yes: compute it, with an explicit model.

When the link is known, compute it with an explicit model, calibrated and validated on data; learning is added only for a demonstrated benefit.

Stories of this step

Principle 2: The model at the core · Next step: Hybridize ↓

Step 2 of 5 · Is the model enough?

Hybridize

Not quite: keep it and learn only what is missing.

When the model falls short, complete it: keep the model, learn only what is missing.

Stories of this step

Principle 2: The model at the core · Next step: Learn frugally ↓

Step 3 of 5 · Are the useful relations known?

Learn frugally

Not well enough: learn from data, with only the complexity that pays.

When the useful relations are not known well enough, evaluate what the data can teach, with only as much complexity as pays.

Stories of this step

Étude MIMO-ELM, figure 1 de la version auteur : mesures et prévisions à une heure pour sept signaux énergétiques corses, environ cent heures en hiver.
Learn frugally · Energy

Seven signals, one ELM

One frugal learner forecasts the whole Corsican electricity mix.

Principle 3: Complexity only if it pays · Next step: AI in its place ↓

Step 4 of 5 · How can people use these tools?

AI in its place

Language models help; the calculation stays with validated codes.

Language models help people use validated tools; the calculation, the checking and the responsibility stay with validated codes and people.

Stories of this step

Principle 4: Every tool, at the periphery · Next step: Demonstrate trust ↓

Step 5 of 5 · How do we know it works?

Demonstrate trust

Compare with a reference and publish code, data and limits.

Judge against a reference, publish code, data and limits, make research findable and reusable.

Stories of this step

Principle 5: Trust is demonstrated

The approach, in detail · All stories, by topic