Open Science Artifact: Agentic Runtime Reconfiguration of Architectural Patterns in Federated Learning

This package contains the results of the experiments presented in the manuscript "Agentic Runtime Reconfiguration of Architectural Patterns in Federated Learning" which is currently under revision.

The tool for replicating the experiments i available at https://github.com/IvanComp/AP4Fed

A .webp screencast on how to run the experiments is available here

Note that the first run may be slower because it involves creating the Docker containers, downloading the training datasets into them, and fetching the LLM using Ollama.

Repository Content:

This repository contains a .zip file with the Federated Learning Agentic Framework. The file consists of:

• README.md: Technical instructions to reproduce the experiments.
• Experiments/: Folder containing all the experiments conducted in the paper (each repeated 10 times) and the Paper Figures.ipynb for analyzing and reproducing the graphs of the paper.
• Experiments/Paper Figures.ipynb: A Jupyter Notebook file used to pre-process and visualize the experimental results based on the output generated by the framework.
• The remaining files are required to run the script that enables Federated Learning Agentic process.

Note: The results of the experiments are inherently non-deterministic and may show minor variations between runs.