Federated-learning simulation

The optional Flower workflow demonstrates SecAgg+ secure aggregation in a single-machine simulation. It is not a validated multi-institutional deployment or a differential-privacy implementation.

The current workflow supports binary classification with leave-one-dataset-out evaluation, one compatible SGDClassifier, and single-table microbiome data.

In an otherwise configured sweep, set:

FEDERATED = mll.Federated(num_rounds=10, local_epochs=2)

Run the separate simulation stage:

mllabiome path/to/config.py --stage federated-evaluate

All study data are loaded on one machine for this demonstration. Real distributed processing requires additional deployment and privacy assessment. See the federated example .