netsci

Analyzing Complex Networks with Python

Author	Version	Demo
Gialdetti

netsci is a python package for efficient statistical analysis of spatially-embedded networks. In addition, it offers several algorithms and implementations (CPU and GPU-based) of motif counting algorithms.

For other models and metrics, we highly recommend using existing and richer tools. Noteworthy packages are the magnificent NetworkX, graph-tool or Brain Connectivity Toolbox.

A simple example

Analyzing a star network (of four nodes)

import numpy as np
import netsci.visualization as nsv

A = np.array([[0,1,1,1], [0,0,0,0], [0,0,0,0], [0,0,0,0]])
nsv.plot_directed_network(A, pos=[[0,0],[-1,1],[1,1],[0,-np.sqrt(2)]])

import netsci.metrics.motifs as nsm
f = nsm.motifs(A, algorithm='brute-force')
print(f)
# [1 0 0 0 0 3 0 0 0 0 0 0 0 0 0 0]

nsv.bar_motifs(f)

GPU speedup

Using GPU for the motif counting is easy

from netsci.models.random import erdos_renyi

# Create an Erdős–Rényi network, and count motifs using a GPU
A_er = erdos_renyi(n=500, p=0.2, random_state=71070)  
f_er = nsm.motifs(A_er, algorithm="gpu")

# Visualize
print(f_er)
# [5447433 8132356 1031546 2023563 1011703 1011109  503098  512458
#   513352  167427   64844  127751   64442   63548   32483    1387]
nsv.bar_motifs(f_er)

The running-time speedup ratio resulting from the GPU-based implementation, as measured over several networks sizes (n) and sparsities (p), is depicted below

A full a live notebook for performing this benmarching is provided below.

Installation

Install latest release version via pip

$ pip install netsci

Install latest development version

via pip

$ pip install git+https://github.com/gialdetti/netsci.git

or in development mode

$ git clone https://github.com/gialdetti/netsci.git
$ cd netsci
$ pip install -e .[dev]

Testing

After installation, you can launch the test suite:

$ pytest

Help and Support

Examples

Theme	MyBinder	Colab
Basic network motifs demo
Connectomics dataset, and 3-neuron motif embedding
Tech: GPU speedup of motif analysis

Communication

Please send any questions you might have about the code and/or the algorithm to eyal.gal@mail.huji.ac.il.

Citation

If you use netsci in a scientific publication, please consider citing the following paper:

Gal, E., Perin, R., Markram, H., London, M., and Segev, I. (2019). Neuron Geometry Underlies a Universal Local Architecture in Neuronal Networks. BioRxiv 656058.

Bibtex entry:

@article {Gal2019
    author = {Gal, Eyal and Perin, Rodrigo and Markram, Henry and London, Michael and Segev, Idan},
    title = {Neuron Geometry Underlies a Universal Local Architecture in Neuronal Networks},
    year = {2019},
    doi = {10.1101/656058},
    journal = {bioRxiv}
}