Data-Driven and Privacy-Preserving Cooperation in Decentralized Learning

This repository stores the csv data generated to produce the results from Data-Driven and Privacy-Preserving Cooperation in Decentralized Learning. The headers mean the following: 'node' ID of the model owner. 'acc' Current accuracy of the model. 'loss' Current loss. 'time' Cumulative time until this epoch. 'epochs' Number of epochs performed. 'n_removed_edges' Number of removed edges from the complete graph. 'type' Type of removal method: random, minimum V distance, minimum dataset distance. 'average_degree' Average degree of the logical topology. 'spectral_gap' Spectral gap of the logical topology. 'data_exchanged' Amount of data that each vertex of one edge have, summed among all edges. '{i}-{j}' Boolean indicating if nodes i and j are connected in the logical topology. 'Vdist_{i}-{j}' Distance between the V matrices of the datasets of i and j. 'Ddist_{i}-{j}' Average L2 distance between the samples of node i dataset and node j dataset. This information can also be found in the next link, please check it for updates: PREDICT 6G / Data-driven-and-privacy-preserving-cooperation-in-decentralized-learning - GitLab (uc3m.es) This work has been partially funded by the European Commission Horizon Europe SNS JU PREDICT-6G (GA 101095890) Project.