This question was migrated from the Slack channel #questions.
Original Question:
I’m doing my MSc thesis. Is Flower using a compression algorithm? I’m doing traffic benchmarking to choose the framework, and I get very low overhead.
Answer:
No, not yet. No compression algo is used. Ig low overhead does not mean compression.
Question 2:
Is Flower using something to reduce traffic, like filtering some layer of the model?
Answer:
No. Only the model parameters you return in client.fit or in fit instructions sent from Strategy