Artifact: Opportunistic Federated Learning: An Exploration of Egocentric Collaboration for Pervasive Computing Applications

In the associated full paper [3], we defined the concept of opportunistic federated learning and proposed various approaches for incorporating nearby devices' data to personalize a deep learning model in an egocentric way. We implemented a simulator using TensorFlow [4] and Keras [2]. The simulator and an associated Docker image are available on GitHub11https://github.com/UT-MPC/swarm and Docker Hub22https://hub.docker.com/repository/docker/sethlee0111/swarm. This artifact can be run in a system that can support Docker engine33https://docs.docker.com/engine/install/. In this artifact guide, we show how the results presented in the paper can be replicated using the Docker image with minimal effort. This guide is also available on the GitHub page.