Module federated 

Federated Split Coordination for Privacy-Preserving ML

This module enables distributed/federated ML workflows where data stays local on each node (sovereignty) and only metadata/sketches cross boundaries.

Architecture

Node A (EU):     data_eu.ald      → local train_eu.ald, test_eu.ald
Node B (US):     data_us.ald      → local train_us.ald, test_us.ald
Node C (APAC):   data_apac.ald    → local train_apac.ald, test_apac.ald
                       ↓
             Coordinator (sees only manifests)
                       ↓
             Global split verification

Structs

FederatedSplitCoordinator

Federated split coordination (no raw data leaves nodes)

GlobalSplitReport

Report on global split quality across all nodes

NodeSplitInstruction

Instructions for a node to execute its split

NodeSplitManifest

Per-node split manifest (shared with coordinator, no raw data)

NodeSummary

Summary for a single node

Enums

FederatedSplitStrategy

Strategy for federated/distributed splitting

SplitQualityIssue

Quality issues that can be detected in federated splits