The detail of algorithm was introduced in our recent paper: "Self-organizing Democratized Learning: Towards Large-scale Distributed Learning Systems", Paper Link
We are working for the demonstration of some properties and initially develop new algorithm for Democratized Learning systems based on our Dem-AI philosophy paper "Distributed and Democratized Learning: Philosophy and Research Challenges", Paper Link
This code is forked and reimplemented based on our prior paper and pFedMe code:
"Federated Learning over Wireless Networks: Convergence Analysis and Resource Allocation" Paper Link
pFedMe implementation as: https://github.com/CharlieDinh/pFedMe
In the current version, we evaluate based on MNIST and Fashion-MNIST Dataset.
Environment: Python 3.8, Pytorch
Downloading dependencies
pip3 install -r requirements.txt
Main File: demlearn_main.py, Setting File: Seting.py, Plot Summary Results: dem_plot_summary.py,
PATH: Output: ./results, Figures: ./figs, Dataset: ./data
Generated Dataset: Google Drive Link, Unzip each dataset file and move to ./data/mnist/data or ./data/fmnist/data
The more detail setting can be found in Setting.py file
- Four algorithms: "fedavg","fedprox","demlearn"
- Datasets: "mnist", "fmnist", "femnist", "Cifar10"
- Clustering metrics: "weight", "cosine"
- Model: "cnn"
- Result file name: "{}_ {}_ I{}_ K{}_ T{}_ b{}_ d{}_ m{}_ {}.h5" that reflects DATASET, RUNNING_ALG, NUM_GLOBAL_ITERS, K_Levels, TREE_UPDATE_PERIOD, beta, DECAY, mu, cluster_metric(w or g).
- E.g.:
fmnist_demlearn_I100_K1_T2_b1-0_dTrue_m0-001_w.h5,mnist_demlearn-p_I60_K1_T60_b1-0_dTrue_m0-002_w.h5
- E.g.:
To run the code with the Setting.py, we use following command:
python demlearn_main.py
To run the plotting results, we use following command:
python dem_plot_summary.py
In particular, we conduct evaluations for specialization (C-SPE) and generalization (C-GEN) of learning models at agents on average that are defined as the performance in their local test data only, and the collective test data from all agents in the region, respectively. Accordingly, we denote Global as global model performance; G-GEN, G-SPE are the average generalization and specialization performance of group models, respectively.
