試験
通常
試験時間: 00:00問題時間: 00:00
問題一覧
問題 58 / 65
If multiple medical institutions want to collaboratively use rare disease data to improve a model without sharing the data, which method is optimal?
①Each organization consolidates raw data on a central server and shares it.
②We use federated learning, where each site trains the model locally and only the parameters are aggregated.
③Each institution publishes all anonymized data and engages in collaborative learning.
④Have each institution train completely independently, and in the end just average the models.
⑤Train only on a single large-scale synthetic dataset and do not distribute it to any institutions.