tags	dataset	framework
basic vision fds	MovieLens	torch torchvision

Federated Recommender System Performance Benchmarking with PyTorch and Flower

This project evaluates the performance of recommender systems that implement collaborative filtering and user policy control in a federated way using the Flower framework. Recommender models use either Matrix Factorization or Variational Autoencoder (VAE), conditioned on item metadata, user preferences, and user policies. Currently the MovieLens dataset is supported. Alternative datasets and implementations of policy-controlled recommender systems will be evaluated in the future.

Set up the project

Install dependencies and project

Install the dependencies defined in pyproject.toml. Simulation has been tested using Python 3.11.6.

pip install -e .

Download MovieLens 1m

Download MovieLens 1m and extract the ml-1m directory to the root project directory alongside the src directory.

Configure Model & Simulation Settings

Edit pyproject.toml.

model-type = <mf|vae>

Run the Project

You can run Flower projects such as this one in both simulation and deployment mode without making changes to the code. Start out by using the simulation mode requires fewer components to be launched manually. By default, flwr run will make use of the Simulation Engine.

Run with the Simulation Engine

flwr run .

You can also override some of the settings for your ClientApp and ServerApp defined in pyproject.toml. For example:

flwr run . --run-config num-server-rounds=5

Run with the Deployment Engine

Note

An update to this example will show how to run this Flower application with the Deployment Engine and TLS certificates, or with Docker.

Visualize the results

Results will be automatically saved to metrics_plots.png and latent_space_visualization.png upon successful run.