Two more, in detail

Every site wants a model shaped to its own patients. A full private copy each is too expensive, and it stops them learning from one another. So the model splits in two. The shared half travels every round. The private half never leaves. Drag the round forward and watch which half is doing the learning.

In plain words: every site keeps a small private piece of the model, shaped to its own patients, and only the shared piece ever travels.

The curve is drawn to show the shape, not to report a run. It is flat while only the shared half learns, then it steps once the private half switches on. That step is the whole point of the method, and you cannot see it in a final number. What is exact here is the architecture: the layer is y = xW + b + scale · GELU(xBA), W is shared and federated, B and A are the private half, and the non linearity between them is what makes it more than a linear adapter. The 3.1% is arithmetic on a 512 by 512 layer. This is the work the ECTRIMS grant was awarded for.

Everything a team needs to start a federated project already exists. It is just scattered across frameworks, legal templates, glossaries and other people's heads, and no two of those speak the same language. Most teams stop at that wall. So we gathered it.

39 pages, eight sections, four life cycle stages and seven worked project stories, covering multiple sclerosis, inflammatory bowel disease, genomics and brain computer interfaces. Built on the model of the ELIXIR research data management kit. Open, and it takes contributions. Pirmani, Vermeulen et al., 2026.