example_weights_matrix: Example Weights Matrix
In adonizetti/BLSM: Bayesian Latent Space Model

"BLSM weights" matrix of a 10 nodes random network for testing purposes

1	example_weights_matrix

A matrix containing positive weights for all pairs of nodes.

Given a couple of nodes, a weight expresses the importance of the distance between the coordinates associated to the two nodes in the latent space in terms of the overall likelihood of the graph. For this reason, even missing links must have a coefficient, otherwise the relative positioning of disconnected nodes would have no effect at all on the graph likelihood.

The exact probability equation is described in BLSM, as well as the notation used.

A few examples:

for unweighted networks, the "BLSM weights" matrix has all the values set to 1.
if two nodes share a strong connection, then the weight coefficient should be greater than 1 so that their positions in the latent space will be closer than they would be in an unweighted framework.
if two nodes share a weak connection, a coefficient smaller than 1 will allow the latent coordinates to be pretty far from each other even though the nodes are connected.

adonizetti/BLSM documentation built on May 25, 2019, 2:24 p.m.