This repository contains the code and experiments for the paper "Signal Recovery from Random Dot-Product Graphs Under Local Differential Privacy".

1. Clone the repository:

$ git clone https://github.com/sidv23/grdpg-ldp.git
$ cd grdpg-ldp
$ julia

2. Activate the Julia environment:

> import Pkg
> Pkg.activate(".")
> Pkg.instantiate()

The notebooks directory contains the Jupyter notebooks for the experiments and simulations. The directory contains the following files:

• clustering.ipynb: The experiment for comparing the clustering accuracy and runtimes for k-Means, SDP clustering, and topology-aware clustering (Experiment 4.3).

• eflip.ipynb: Illustration of how edgeFlip results in a transformation of the latent configuration underlying GRDPGs (Figure 2).

• flights.ipynb: Illustration of recovering latent positions using UMAP for the under $\epsilon$-LDP (Experiment 4.1).

• pd_convergence.ipynb: Experiment to illustrate recovery of topological information under edgeFlip (Experiment 4.2).

• heatmap.ipynb: Heatmap of $d_{2, \infty}$ error for the different values of $\epsilon$ and $n$ (Experiment 3.1).

• top-clustering.ipynb: Topology-aware spectral clustering with more details.

The src directory contains the .jl source-code and the data directory contains the data files.

For other issues, please open a new issue.

@inproceedings{vishwanath2025signal,
  title={Signal Recovery from Random Dot-Product Graphs under Local Differential Privacy},
  author={Vishwanath, Siddharth and Hehir, Jonathan},
  booktitle={International Conference on Artificial Intelligence and Statistics},
  year={2025},
  organization={PMLR}
}