README.md

In rcannood/incgraph: Incremental Graphlet Counting for Network Optimisation

incgraph

incgraph incrementally calculates the differences in orbit counts when performing single edge modifications in a network. For evolving networks, calculating the differences in orbit counts is much more performant than recalculating all orbit counts from scratch for each time point.

Install

incgraph currently needs to be installed using devtools. When building for Windows, the correct version of Rtools also needs to be installed.

devtools::install_github("rcannood/incgraph")

Examples

Create a new (empty) network with 4 nodes

library(incgraph)
net <- new.incgraph.network(amnt.nodes = 4)

Create a new network with 4 nodes and some edges

net <- new.incgraph.network(links = matrix(c(1, 2, 2, 3, 1, 4), ncol=2))

Create a new network with 10 nodes and some edges

net <- new.incgraph.network(amnt.nodes = 10, links = matrix(c(1, 2, 2, 3, 1, 4), ncol=2))

Create a more complex network from a matrix

mat <- matrix(c(1, 2, 
                1, 3, 
                1, 4, 
                1, 5, 
                1, 6, 
                1, 7, 
                2, 7, 
                2, 8, 
                2, 9, 
                2, 10), ncol=2)
net <- new.incgraph.network(links=mat)

Calculate the initial orbit counts using orca

orb.counts <- calculate.orbit.counts(net)

Modify an edge and calculate the differences in orbit counts

flip(net, 5, 10) # add (5,10)
delta1 <- calculate.delta(net, 5, 10)

Modify another edge

flip(net, 6, 10) # add (6, 10)
delta2 <- calculate.delta(net, 6, 10)

And another

flip(net, 1, 5)  # remove (1, 5)
delta3 <- calculate.delta(net, 1, 5)

Verify that the new orbit counts equals the old orbit counts plus the delta counts

new.orb.counts.incremental <- orb.counts + 
  delta1$add - delta1$rem +
  delta2$add - delta2$rem +
  delta3$add - delta3$rem
new.orb.counts <- calculate.orbit.counts(net)
all(new.orb.counts.incremental == new.orb.counts) # TRUE

## [1] TRUE

Examples of additional helper functions

Transform the network to a matrix

network.as.matrix(net)

##      [,1] [,2]
## [1,]    1    2
## [2,]    1    5
## [3,]    2    7
## [4,]    3    7
## [5,]    4    8
## [6,]    5    9
## [7,]    5   10

Get all neighbours of a node

get.neighbours(net, 1)

## [1] 0 1 4

Does the network contain a specific interaction?

contains(net, 5, 10)

## [1] TRUE

contains(net, 7, 10)

## [1] FALSE

Reinitialise to an empty network

reset(net)
network.as.matrix(net)

##      [,1] [,2]

rcannood/incgraph documentation built on Dec. 11, 2019, 12:25 p.m.