This repository contains the code for the paper "Neural Blockmodeling for Multilayer Networks" by Marcin Pietrasik and Marek Reformat.
First, unpack the datasets using:
unzip datasets.zip
The code for our method was ran using Python version 3.6.12 along with the packages listed in requirements.txt. To install these packages, run:
pip install -r requirements.txt
main.py takes three optional command line arguments:
| Parameter | Default | Description |
|---|---|---|
| -h, --help | False | shows help message and exits |
| -d DATASET, --dataset DATASET | trade | name of dataset in datasets directory |
| -a, --A | 8 | integer value of the A hyperparameter |
| -e, --E | 8 | integer value of the E hyperparameter |
| -k, --K | 4 | integer value of the K hyperparameter |
| -t TASK, --task TASK | link_prediction | task to be performed, choosen from: link_prediction, node_classification, community_detection. |
| -s SPLIT, --split SPLIT | 0.8 | float value of training split size |
| -p EPOCHS, --epochs EPOCHS | 1000 | integer value of number of epochs |
| -v, --save | False | flag indicating whether to save trained model, embeddings, and community memberships |
The commands for running our model using hyperparameters used in the paper are found in commands.txt