some qustions about oneshot_matching experiment

[leondada]

Hi, i have something that confused me:

1. What is the difference between oneshot_matching and BBP_MAP?
2. The retrain process in matching actually introduces multiple original data information, so does the matching really reflect his ability to aggregation in a single communication?

[leondada]

and in FL, data privacy is very important， how to do this retrain without increasing the number of communications?