Python implementation of Dynamic SpringRank model described in:

• A model for efficient dynamical ranking in networks
  Andrea Della Vecchia, Kibidi Neocosmos, Daniel B. Larremore, Cristopher Moore, and Caterina De Bacco.
  Physical Review E, 110, 034310 (2024).

• A preprint is also available.

Dynamic SpringRank is a physics-inspired method for inferring dynamic rankings in directed temporal networks — networks in which each directed and timestamped edge reflects the outcome and timing of a pairwise interaction. It is the natural extension of SpringRank to temporal networks.

• src : contains the Python implementation of two versions of Dynamic SpringRank (Online and Offline) as well as SpringRank.
• data/input : contains synthetic data used to illustrate the functioning of the models. Note: 'static' data in folder refers to data without a meaningful time component (refer to 'The Relevance of Time' section in paper for further details)
• data/output : location for saved results after running models.

The project was developed in Python 3.8.11 with the packages contained in the requirements.txt. We recommend creating a conda environment and installing the pre-requisite packages with conda create --name DSR --file requirements.txt

The model can be run by executing the run_models.py while in the src directory. The run_models.py script has the following arguments:

• --model select the model to run
• --dataset file name of the synthetic dataset that will be used (excluding file extension)
• --save flag that saves the output of the models
• --verbose flag that prints more details of internal procedures of model as it is running

The model accepts as input a numpy array of shape [T, N, N], where T is the number of timesteps and N is the number of nodes in the network.

The model outputs a Python Tuple containing the evaluations metrics and rankings of the nodes. Note: when the --save flag is called, the output is saved as a Python Dictionary containing the aforementioned results as well as the runtime of the model.