Artifact: Prototype for Collaborative Machine Learning in Federated Data Spaces

This artifact provides a proof-of-concept prototype accompanying the ICSA 2026 paper “Architectural Foundations for Collaborative Machine Learning in Federated Data Spaces.” It implements a simplified coordination workflow for cross-domain Federated Learning (FL) on top of a federated broker layer (MQTT with NGSI-LD-style entities). The artifact demonstrates: (i) publishing an FL job as a discoverable NGSI-LD entity, (ii) propagating this job across a small broker federation, (iii) client-side discovery via subscription, and (iv) executing FL rounds using a standard FL runtime. The prototype supports a single-machine “local federation” for functional verification and a multi-machine deployment for reproducing the paper’s experimental setup style (separate brokers/clients/servers). Documentation includes step-by-step broker startup, subscription-based discovery, example commands for running a minimal end-to-end FL job, and instructions for replicating paper figures.