In Ecological-Complexity-Lab/infomap_ecology_package: Community Detection using Infomap, Inspired by Ecological Networks

# Use this to add a copy-code buttong
klippy::klippy(position = c('top', 'right'), color = 'darkred')

knitr::opts_chunk$set(echo = TRUE)

library(emln)
library(infomapecology)

Data set

Load the 65th multilayer network from the emln package. This particular network is a temporal plant pollinator multilayer network without interlayer interactions, and it consists of 6 layers, corresponding to the flowering season between 2006 and 2011.

library(emln)
# Load multilayer network number 65
emln_65 <- load_emln(65)
emln_65$layers
head(emln_65$nodes)
head(emln_65$extended)

Input

A multilayer link-list format. With this format, a random walker moves within a layer and with a given relax rate r jumps to another layer without recording this movement, such that the constraints from moving in different layers can be gradually relaxed. If the *Inter exists, then interlayer edges can be used. However, if a relax rate is also used (Infomap input parameter --multilayer-relax-rate), the interlayer edges are ignored. Here we only include *Intra.

# A network in multilayer format
*Intra
#layer node node weight
1 101 1 19
1 113 1 2
1 152 1 1
... some more cases ...
6 69 11 17
6 70 11 2
6 87 11 3

R Code

The description of functions create_multilayer_object and run_infomap_multilayer in the infomapecology package contains everything you need to know. The for loop performs a sensitivity analysis to examine how structure changes with increasing relax rates.

#Run Infomap and return the network's modular structure at increasing relax-rates.
relaxrate_modules <- NULL
# Initialize an empty data frame to store results
results <- data.frame(relax_rate = numeric(), module_count = numeric())

for (r in seq(0.001,1, by = 0.0999)){
  print(r)
  modules_relax_rate <- run_infomap_multilayer(emln_65, relax = T, silent = T, trials = 50, seed = 497294,    multilayer_relax_rate = r, multilayer_relax_limit_up = 1, multilayer_relax_limit_down = 0, temporal_network = T)
  modules_relax_rate$modules$relax_rate <- r
  relaxrate_modules <- rbind(relaxrate_modules, modules_relax_rate$modules)
  # Append results to the data frame
  results <- rbind(results, data.frame(relax_rate = r, module_count = modules_relax_rate$m))

}

ggplot(results, aes(x = relax_rate, y = module_count)) +
  geom_line(color = "blue") +
  labs(x = "relax rate (r)", y = "Number of modules")+
  theme_bw()+theme(panel.grid.minor = element_blank())

Infomap

Under the hood, the function run_infomap_multilayer runs:

./Infomap infomap_multilayer.txt . -2 --seed 497294 -N 50 -i multilayer --multilayer-relax-rate 0.15 --multilayer-relax-limit-up 1 --multilayer-relax-limit-down 0 --silent

Explanation of key arguments: -i multilayer indicates a multilayer input format, which is automatically recognized as a multilayer link list (and not as a general link list). --multilayer-relax-rate 0.15 defines the relax rate (here 0.15). * multilayer-relax-limit-up 1 --multilayer-relax-limit-down 0 limits relaxation to a single layer upwards (l to l+1), but never downwards, because time flows one-way.

Output

For multilayer network the output file has a _states suffix, with the following format. This output is as in Temporal multilayer network with interlayer edges.

# path flow name stateId physicalId layerId
# path flow name state_id node_id layer_id
1:1 0.0208474 "12" 604 12 6
1:2 0.0166813 "100" 531 100 6
1:3 0.0103777 "45" 600 45 6
1:4 0.00927889 "58" 526 58 6

This output is parsed by run_infomap_multilayer to obtain a table in which each state node (combination of a physical node in a layer) is assigned to a module. This can be obtained by (after running the code above):

#relaxrate_modules$modules
relaxrate_modules %>% select(node_id, layer_id, node_name, type, module, relax_rate)

Ecological-Complexity-Lab/infomap_ecology_package documentation built on June 6, 2024, 5:28 a.m.