graph-ensembles 0.2.0

The graph ensemble package contains a set of methods to build fitness based graph ensembles from marginal information.

Graph ensembles

The graph ensemble package contains a set of methods to build fitness based graph ensembles from marginal information. These methods can be used to build randomized ensembles preserving the marginal information provided.

• Free software: GNU General Public License v3

• Documentation: https://graph-ensembles.readthedocs.io.

Installation

Install using:

pip install graph_ensembles

Usage

Currently only the RandomGraph and StripeFitnessModel are fully implemented. An example of how it can be used is the following. For more see the example notebooks in the examples folder.

import graph_ensembles as ge
import pandas as pd

v = pd.DataFrame([['ING', 'NL'],
                 ['ABN', 'NL'],
                 ['BNP', 'FR'],
                 ['BNP', 'IT']],
                 columns=['name', 'country'])

e = pd.DataFrame([['ING', 'NL', 'ABN', 'NL', 1e6, 'interbank', False],
                 ['BNP', 'FR', 'ABN', 'NL', 2.3e7, 'external', False],
                 ['BNP', 'IT', 'ABN', 'NL', 7e5, 'interbank', True],
                 ['BNP', 'IT', 'ABN', 'NL', 3e3, 'interbank', False],
                 ['ABN', 'NL', 'BNP', 'FR', 1e4, 'interbank', False],
                 ['ABN', 'NL', 'ING', 'NL', 4e5, 'external', True]],
                 columns=['creditor', 'c_country',
                          'debtor', 'd_country',
                          'value', 'type', 'EUR'])

g = ge.Graph(v, e, v_id=['name', 'country'],
             src=['creditor', 'c_country'],
             dst=['debtor', 'd_country'],
             edge_label=['type', 'EUR'],
             weight='value')

# Initialize model
model = ge.StripeFitnessModel(g)

# Fit model parameters
model.fit()

# Sample from the ensemble
model.sample()

Development

Please work on a feature branch and create a pull request to the development branch. If necessary to merge manually do so without fast forward:

git merge --no-ff myfeature

To build a development environment run:

python3 -m venv env
source env/bin/activate
pip install -e '.[dev]'

For testing:

pytest --cov

Credits

This is a project by Leonardo Niccolò Ialongo and Emiliano Marchese, under the supervision of Diego Garlaschelli.

History

0.2.0 (2021-07-12)

• Added likelihood and nearest neighbour properties.

• Revisited API for measures to ensure correct recompute if necessary.

0.1.3 (2021-04-29)

• Added new option for fitting the stripe model that ensures that the minimum non-zero expected degree is one

• Corrected issue in expected degree calculations

0.1.2 (2021-04-07)

• Added scale invariant probability functional to all models

• Improved methods for convergence with change in API, xtol now a relative measure

• Added pagerank and trophic depth to the library

• Added methods for graph conversion to networkx

• Added methods for computing the adjacency matrix as a sparse matrix

0.1.1 (2021-03-29)

• Fixed bug in stripe expected degree computation

• Added testing of expected degree performance

0.1.0 (2021-03-29)

• Added the block model and group info to graphs

• Added fast implementation of theoretical expected degrees

• Fixed some compatibility issues with multiple item assignments

0.0.4 (2021-03-15)

• Fixed issues with slow pandas index conversion

0.0.3 (2021-03-14)

• Large changes in API with great improvements in usability

• Added sampling function

• Added RandomGraph model

• Added Graph classes for ease of use

0.0.2 (2020-11-13)

• Added steps for CI.

• Corrected broken links.

• Removed support for python 3.5 and 3.6

0.0.1 (2020-10-28)

• First release on PyPI. StripeFitnessModel available, all other model classes still dummies.