This directory contains reproducible example scripts demonstrating core Unbitrium functionality.
Minimal end-to-end federated learning simulation using FedAvg aggregation on synthetically partitioned data.
python examples/quickstart.py
Demonstrates various non-IID partitioning strategies including Dirichlet, Mixture-of-Dirichlet-Multinomials (MoDM), quantity skew, and entropy-controlled partitioning. Generates visualization PNG files.
python examples/dirichlet_partitioning.py
Compares FedAvg, FedSim, and pFedSim aggregation strategies under heterogeneous data distributions.
python examples/similarity_aggregation.py
Full benchmark harness demonstration with experiment provenance tracking and standardized result artifacts.
python examples/benchmark_run.py
All examples require the Unbitrium package to be installed:
pip install unbitrium
Or install from source:
pip install -e .