version 0.8 federation-learning example seems contain an error in validation dataset setting about tf-encrypted

[zjn-code]

Yes, this is a mistake, and DataOwners shouldn't use same train dataset.

data_owners = [
        DataOwner(
            "data-owner-0", "./data/train.tfrecord", model_owner.build_update_step
        ),
        DataOwner(
            "data-owner-1", "./data/train.tfrecord", model_owner.build_update_step
        ),
        DataOwner(
            "data-owner-2", "./data/train.tfrecord", model_owner.build_update_step
        ),
    ]

Our new version 0.9.0 fix this problem, you could check this

[jvmncs]

fwiw this example was meant to be a demonstration of just FL logic/orchestration, and since the data is fake/generated I think we didn't really care to do the example properly at the time.

in practice, you might want each data owner to partition a hold-out validation set from each of their training sets, or perhaps include an entirely new/separate validation set from the model owner. you definitely wouldn't want to create the validation set from your test set, otherwise any kind of hyperparameter tuning would bleed into your testing and lead to overly optimistic results