Federated Genetic Algorithm

Federated Learning simulation using Flower where each client runs a Genetic Algorithm to optimize hyperparameters locally before training a CNN on CIFAR-10 / MNIST.

Project structure

.
├── data/                        # CIFAR-10 and MNIST datasets (local)
├── federatedgeneticalgorithm/   # Flower app (server + clients)
│   ├── federatedgeneticalgorithm/
│   │   ├── client_app.py        # ClientApp — GA + local training
│   │   ├── server_app.py        # ServerApp — aggregation strategy
│   │   ├── genetic_algorithm.py # DEAP-based GA for HP search
│   │   ├── surrogate_model.py   # RandomForest surrogate for fitness estimation
│   │   ├── task.py              # Model definition, train/test helpers
│   │   └── config/config.py     # Global config (device, seed, …)
│   ├── scripts/
│   │   ├── run_federated_learning_cpu.sh
│   │   └── run_federated_learning_gpu.sh
│   └── pyproject.toml           # Flower app config & dependencies
├── notebooks/                   # Exploratory notebooks
└── pyproject.toml               # Root workspace dependencies

Requirements

• Python 3.10
• uv (recommended) or pip

Installation

# 1. Create and activate a virtual environment
uv venv --python 3.10
source .venv/bin/activate   # macOS / Linux
# .venv\Scripts\activate    # Windows

# 2. Install the Flower app and its dependencies
cd federatedgeneticalgorithm
pip install -e .

Running

All commands below must be run from inside the federatedgeneticalgorithm/ directory.

CPU (default)

flwr run .

GPU

flwr run . local-simulation-gpu

Alternatively, use the provided scripts from the project root:

bash federatedgeneticalgorithm/scripts/run_federated_learning_cpu.sh
bash federatedgeneticalgorithm/scripts/run_federated_learning_gpu.sh

Tuning the run

Key hyperparameters live in federatedgeneticalgorithm/pyproject.toml under [tool.flwr.app.config]:

Key	Default	Description
num-server-rounds	20	Number of federated rounds
fraction-train	0.5	Fraction of clients selected/round
local-epochs	5	Local training epochs per client

Override any value at runtime without editing the file:

flwr run . --run-config "num-server-rounds=10 local-epochs=3"