in the paper 'Swarm Learning for decentralized and confidential clinical machine learning',
method - parameter merging section, the weighted average is defined as
Pm = sum(Wi Pi) / (n sum(Wi))
and 'n' represents number of nodes.
May i ask why 'n' is in the denominator? Is there any special concern here?
I think it should be *Pm = sum(Wi Pi) / sum(Wi). when take equal weights, it becomes the simple average without weight Pm = sum(Pi) / sum(1) = sum(Pi) / n** ？
Hi,
in the paper 'Swarm Learning for decentralized and confidential clinical machine learning', method - parameter merging section, the weighted average is defined as
Pm = sum(Wi Pi) / (n sum(Wi)) and 'n' represents number of nodes.
May i ask why 'n' is in the denominator? Is there any special concern here?
I think it should be *Pm = sum(Wi Pi) / sum(Wi). when take equal weights, it becomes the simple average without weight Pm = sum(Pi) / sum(1) = sum(Pi) / n** ？