Federated Submodular Maximization with Differential Privacy

This is for releasing the source code of the paper "Federated Submodular Maximization with Differential Privacy".

Requirements

• Python==3.8
• numpy==1.22.3

How to use

To reproduce the experiments, do:

1. download dataset

You can download dataset in here, and put it in the same folder as run.py, such as code-dpsm/data/DBLP

2. run

Run the script run.py:

$ python3 run.py

Before you run the script, you can make some modifications to it.

Such as in line 38:

H = Handler(MaxP=31, save_path="res.csv")

MaxP=31 means the maximum number of parallel computing processes is 31

save_path="res.csv" means experiment results will be saved in res.csv

After the program has finished running, you need to use res_reader.py to process the data and the final results will be saved in the corresponding out_*.csv file