Python implementation of the Latent Space Hawkes (LSH) model for continuous-time dynamic networks
This repo contains the Python implementation and code to replicate experiments for LSH model in the paper A Mutually Exciting Latent Space Hawkes Process Model for Continuous-time Networks, accpeted by UAI 2022.
Networks and temporal point processes serve as fundamental building blocks for modeling complex dynamic relational data in various domains. We propose the latent space Hawkes (LSH) model, a novel generative model for continuous-time networks of relational events, using a latent space representation for nodes. We model relational events between nodes using mutually exciting Hawkes processes with baseline intensities dependent upon the distances between the nodes in the latent space and sender and receiver specific effects. We propose an alternating minimization algorithm to jointly estimate the latent positions of the nodes and other model parameters. We demonstrate that our proposed LSH model can replicate many features observed in real temporal networks including reciprocity and transitivity, while also achieves superior prediction accuracy and provides more interpretability compared to existing models.
Dependencies (with python >= 3.7)
- numpy==1.18.5
- scikit-learn==0.23.1
- matplotlib==3.2.2
- networkx==2.4
- dynetworkx==0.3.2
- seaborn==0.10.1
- scipy==1.5.0
- autograd=1.3
- pickle=4.0
enron-events.pckl: stores the Enron data provided in Yang's github-repo. All other datasets are available in the storage/datasets; All results will be stored in the storage/results
It will run the estimation and save estimated parameters in a pickle file.
# run the model on reality mining dataset with a 2 dimendion LSH model
python LSH_model_fit.py --data 'reality' --dim 2This task requires running the future events prediction tasks (model estimation) first to load the estimated parameters from the pickle file.
# run dynamic link prediction on reality mining dataset with a 2 dimendion model
python dynamic_link_pred.py --data 'reality' --dim 2This task requires running the future events prediction tasks (model estimation) first to load the estimated parameters from the pickle file.
# run the generative test on reality mining dataset with a 2 dimendion model
python LSH_generative.py --data 'reality' --dim 2# run the simulation experiments
python LSH_model_sim.pyoptional arguments:
--data Dataset to use (eg. reality, Enron-Yang, MID, fb-forum)
-d --dim Latent dimensions for LSH model
-c --continent Whether to plot 2D MID latent space with continent colored
-a --alternating Whether to use alternating minimization for latent space model estimation
-n --negative Whether to use negative slope for latent space model