BiCM documentation¶
This package provides a tool for calculating the Bipartite Configuration Model (BiCM), as introduced in:
`Fast and scalable likelihood maximization for Exponential Random Graph Models with local constraints. Vallarano et al., Scientific Reports 11.1 (2021) <https://doi.org/10.1038/s41598-021-93830-4>`_,
`Randomizing bipartite networks: the case of the World Trade Web. Saracco et al., Sci Rep 5, 10595 (2015). <https://doi.org/10.1038/srep10595>`_
The monopartite projection on one layer is introduced in:
`Inferring monopartite projections of bipartite networks: an entropy-based approach, Saracco et al., 2017 New J. Phys. 19 053022 <https://doi.org/10.1088/1367-2630/aa6b38>`_ .
The weighted model BiWCM is introduced in:
`Inferring comparative advantage via entropy maximization, Bruno et al., 2023 Journal of Physics: Complexity, Volume 4, Number 4 <https://doi.org/10.1088/2632-072X/ad1411>`_ .
Basic functionalities¶
To install:
pip install bicm
To import the module:
import bicm
To generate a Graph object and initialize it (with a biadjacency matrix, edgelist or degree sequences):
from bicm import BipartiteGraph
myGraph = BipartiteGraph()
myGraph.set_biadjacency_matrix(my_biadjacency_matrix)
myGraph.set_adjacency_list(my_adjacency_list)
myGraph.set_edgelist(my_edgelist)
myGraph.set_degree_sequences((first_degree_sequence, second_degree_sequence))
Or alternatively, with the respective data structure as input:
from bicm import BipartiteGraph
myGraph = BipartiteGraph(biadjacency=my_biadjacency_matrix, adjacency_list=my_adjacency_list, edgelist=my_edgelist, degree_sequences=((first_degree_sequence, second_degree_sequence)))
To compute the BiCM probability matrix of the graph or the relative fitnesses coefficients as dictionaries containing the nodes names as keys:
my_probability_matrix = myGraph.get_bicm_matrix()
my_x, my_y = myGraph.get_bicm_fitnesses()
This will solve the bicm, or the relative biwcm model if the graph is weighted, using recommended settings for the solver. To customize the solver you can alternatively use (in advance) the following method:
myGraph.solve_tool(light_mode=False, method='newton', initial_guess=None, tolerance=1e-8, max_steps=None, verbose=False, linsearch=True, regularise=False, print_error=True, exp=False)
You can compute the p-value of each (weighted) link in the network. To return the matrix of pvalues or the binarisation after applying a threshold:
myGraph.get_weighted_pvals_mat()
myGraph.get_validated_matrix(significance=0.01, validation_method='fdr')
To get the rows or columns projection of the graph:
myGraph.get_rows_projection()
myGraph.get_cols_projection()
Alternatively, to customize the projection:
myGraph.compute_projection(rows=True, alpha=0.05, method='poisson', threads_num=4, progress_bar=True)
Dependencies¶
This package has been developed for Python 3.5 but it should be compatible with other versions. It works on numpy arrays and it uses numba and multiprocessing to speed up the computation. Feel free to send an inquiry (and please do it!) if you find any incompatibility.