networkRobustness

Robustness profiles and visualization for complex networks.
Simulate static node failures (random or targeted by centrality/entanglement), aggregate across many network realizations, and produce ribbons + critical-fraction bar plots.

---

Repository layout

R/
  critical_points.R            # pc computation (critical fractions)
  Graphics.R                   # palettes, multiplot, helpers
  io_profiles.R                # read/write helpers for robustness profiles
  labels.R                     # model/measure pretty names
  networkRobustness-package.R  # package docs
  plots.R                      # plotting helpers (ribbons, bars)
  Robustness.R                 # core robustness simulation API
  zzz.R                        # package init

inst/scripts/
  batch_entanglement_profiles.R          # build entanglement robustness from MI profiles
  build_profiles.R                        # aggregate+plot standard measures
  build_profiles_with_entanglement.R      # aggregate+plot (standard + entanglement)
  generate_proxy_MI_profiles.R            # optional: (re)generate MI profiles (proxy)
  generate_synthetic_networks_and_profiles.R # generate synthetic networks + standard profiles

---

Installation

# from a clean R session
install.packages(c(
  "igraph","ggplot2","ggraph","gplots","RColorBrewer","ggsci"
))

# install devtools if needed
if (!requireNamespace("devtools", quietly = TRUE)) install.packages("devtools")

# install this package from your local checkout
devtools::document()   # generate Rd docs
devtools::install()    # install networkRobustness
devtools::check()      # optional: run R CMD check (should pass cleanly)

# run unit tests (optional)
devtools::test()

Note: We target igraph >= 2.0. The code already uses the non-deprecated API (e.g., delete_vertices(), components()).

---

Data directories expected by the scripts

Scripts assume a project root working directory with the following (created or to be created):

<project root>/
  ER_N256/       BA_N256/      WS_N256/      RGG_N256/      SBM4_N256/      HSBM4_N256/
    <MODEL>_r<ID>.edges             # undirected edgelist, 0-based node ids
    <MODEL>_r<ID>_robustness.csv    # (standard measures; created by generate_* script)
  Entanglement/
    <MODEL>_N256/
      <MODEL>_r<ID>_MI-profile.csv[.gz]    # MI profiles (inputs for entanglement)
      <MODEL>_r<ID>_robustness.csv         # entanglement robustness (output)

• Edgelists in <MODEL>_N256/ use 0-based node labels; scripts shift them to 1..N internally.
• MI profiles may be gzipped (.csv.gz). Scripts accept both .csv and .csv.gz.

---

What each script does (and in what order to run)

0) Quick glossary

• Standard measures: degree, betweenness, PageRank, eigenvector, coreness, clustering, plus random.
• Entanglement: the MI-based ranking read from *_MI-profile.csv(.gz) (we pick the β* that minimizes mean entropy).

---

1) Generate synthetic networks + standard robustness (optional)

Rscript "$(Rscript -e 'cat(system.file("scripts","generate_synthetic_networks_and_profiles.R", package="networkRobustness"))')"

Inputs: none (networks are generated).
Outputs: in each <MODEL>_N256/ folder: - <MODEL>_r<ID>.edges - <MODEL>_r<ID>_robustness.csv (standard measures only)

If you already have <MODEL>_N256/*.edges and standard *_robustness.csv, you can skip this step.

---

2) Provide (or generate) MI entanglement profiles

If you already have MI profiles: place them under Entanglement/<MODEL>_N256/ as <MODEL>_r<ID>_MI-profile.csv or <MODEL>_r<ID>_MI-profile.csv.gz.

If you need to (re)generate proxy MI profiles:

Rscript "$(Rscript -e 'cat(system.file("scripts","generate_proxy_MI_profiles.R", package="networkRobustness"))')"

Outputs: Entanglement/<MODEL>_N256/<MODEL>_r<ID>_MI-profile.csv.gz (or .csv).

---

3) Build entanglement robustness from MI profiles

Rscript "$(Rscript -e 'cat(system.file("scripts","batch_entanglement_profiles.R", package="networkRobustness"))')"

Inputs: - Entanglement/<MODEL>_N256/<MODEL>_r<ID>_MI-profile.csv[.gz] - <MODEL>_N256/<MODEL>_r<ID>.edges

Outputs: - Entanglement/<MODEL>_N256/<MODEL>_r<ID>_robustness.csv (with centr = "ent")

---

4) Aggregate + plot (standard measures only)

Rscript "$(Rscript -e 'cat(system.file("scripts","build_profiles.R", package="networkRobustness"))')"

Outputs: synth_robustness_standard.png

---

5) Aggregate + plot (standard + entanglement), legacy style

Rscript "$(Rscript -e 'cat(system.file("scripts","build_profiles_with_entanglement.R", package="networkRobustness"))')"

Outputs: synth_robustness_with_bars.png

---

Troubleshooting

• Ensure files are named <MODEL>_r<ID>_MI-profile.csv or .csv.gz exactly.\
• Check robustness CSVs have a centr column (older may use attack).\
• Missing entanglement = missing robustness files in Entanglement/….\
• Differences in RGG/SBM curves usually mean different β* selected.

---

Minimal R example

library(igraph)
library(networkRobustness)

set.seed(1)
g <- sample_gnp(100, 0.05)

deg <- igraph::degree(g)
prof_deg <- getStaticFailureProfile(g, deg, attack = "deg")
prof_rnd <- getStaticFailureProfile(g, NULL, attack = "random", MC = 10)

head(prof_deg)
head(prof_rnd)

---

License

MIT © Authors.

Issues: https://github.com/CoMuNeLab/RobustnessProfiles/issues