Collection of networks

Collection of networks#

The library provides a basic set of networks which can be read directly. Also it contains a generator of multilayer networks according to Erdos=Renyi and Preferential attachement models.

Networks shipped with the library#

A toy network used with by the authors of CIM method.

get_toy_network_cim() → MultilayerNetwork#

Get a toy network that was used by the authors of CIM method.

The paper is here: https://doi.org/10.1007/s10489-021-02656-0

A three-layered network with 11 actors made up by Piotr.

get_toy_network_piotr() → MultilayerNetwork#

Get threelayered toy network easy to visualise.

Loader of the l2_course_net network.

get_l2_course_net(node_features: bool, edge_features: bool, directed: bool) → TemporalNetwork#

Read l2_course_net.

A network built from interactions between students during the three months long, abroad language course of the Arabic untertaken by 41 US students. This funciton loads a TemporalNetwork with three snapshots, each with two layers: “ego” (with edges picked as top 5 peers by each student) and “course” (with edges obtained by filtering out a complete graph according to a monthly questionarre).

For details see a paper which analyses the network: “Peer Interaction Dynamics and Second Language Learning Trajectories During Study Abroad: A Longitudinal Investigation Using Dynamic Computational Social Network Analysis” by M.B. Paradowski, N. Whitby, M. Czuba, and P. Bródka in “Language Learning”, 2024 (https://doi.org/10.1111/lang.12681). Please note that the networ used in paper was slightly different constructed: (1) it was a singlelayer network, (2) nodes 11 and 17 were discarded due to uncertainity of their data, (3) sources were not anonymised.

Parameters:

  • node_features – wether to to load features of the nodes (mainly language profficiency and psychological metrics with personal data).

  • edge_features – wether to load features of the edges (intensity of contact and usage of the second language)

  • directed – wether to load the network as a directed graph; if so nummerical attributes of edges between corresponding nodes are averaged while other attributes are discarded

Returns:

a temporal, multilayer network

Generators of artificial networks#

class MultilayerERGenerator(nb_layers: int, nb_actors: int, nb_steps: int, std_nodes: int)#

Bases: MultilayerBaseGenerator

Erdos-Renyi multilayer networks generator.

__init__(nb_layers: int, nb_actors: int, nb_steps: int, std_nodes: int) → None#

Initialise the object.

Parameters:

  • nb_layers – number of layers of the generated network

  • nb_actors – number of actors in the network

  • nb_steps – number of steps of the generative algorithm which builds the network

  • std_nodes – standard deviation of the number of nodes in each layer (expected value is a number of actors)

get_models() → list[PyEvolutionModel]#

Get the the evolutionary model.

For each layer with num nodes drawn from a standard distribution.

Returns:

list of Erdos-Renyi generators

class MultilayerPAGenerator(nb_layers: int, nb_actors: int, nb_steps: int, nb_hubs: int)#

Bases: MultilayerBaseGenerator

Preferential Attachment multilayer networks generator.

__init__(nb_layers: int, nb_actors: int, nb_steps: int, nb_hubs: int) → None#

Initialise the object.

Parameters:

  • nb_layers – number of layers of the generated network

  • nb_actors – number of actors in the network

  • nb_steps – number of steps of the generative algorithm which builds the network

  • nb_seeds – number of seeds in each layer and a number of egdes from each new vertex

get_models() → list[PyEvolutionModel]#

Get evolutionary models for each layer.

Returns:

list of Preferential Attachment generators

generate(model: Literal['PA', 'ER'], nb_actors: int, nb_layers: int) → MultilayerNetwork#

Generate a multilayer Erdos-Renyi or Preferential-Attachement net.

Contents