This repo contains a parallel implementation of Kiran Garimella's algorithm to compute the random walk controversy score of a graph.
Let G(V,E) be a graph with two partitions X and Y; in the paper
"Quantifying Controversy on Social Media" Garimella et al. define the
Random Walk Controversy (RWC) measure as follows: "Consider two random walks, one ending in partition X and one ending
in partition Y , RWC is the difference of the probabilities of two events: (i) both random walks started from the
partition they ended in and (ii) both random walks started in a partition other than the one they ended in.”.
The measure is quantified as
Since the probabilities are computed by making simulations consisting in random walks and the simulations are independent
of each other, they can easily be done in parallel.
The following table shows the performance and results comparison made between the (sequential)
implementation provided by one of the original authors and this
implementation. The datasets can be found in the same repo as the original authors' implementation.
| Dataset | Seq. time (s) | Seq. RWC score | Par. time (s) | Par. RWC score | Speedup |
|---|---|---|---|---|---|
| baltimore | 618 | 0.869 | 71 | 0.872 | 8.70 |
| beefban | 128 | 0.873 | 19 | 0.882 | 6.73 |
| gunsense | 2232 | 0.851 | 221 | 0.853 | 10.10 |
| indiana | 229 | 0.720 | 37 | 0.727 | 6.19 |
| indiasdaughter | 490 | 0.825 | 62 | 0.832 | 7.90 |
Beside the evident speedup obtained by exploiting a multicore architecture, we can see that the sequential and parallel versions almost converge to the same results.
To install the latest version of the library just download (or clone) the current project, open a terminal and run the following commands:
pip install -r requirements.txt
pip install .
Alternatively use pip
pip install random-walk-controversy
python3 -m random_walk_controversy [-h] [-v] [-l] edgelist community1_nodelist community2_nodelist percent n
More info about the parameters can be fetched by using the -h option.
The option -v can be used to increase output verbosity and print, alongside the rwc score, the statistics about
random walks. If not specified, only the RWC score is printed out.
Finally, the -l option displays a log on the terminal everytime a simulation is completed; I have found this option
pretty usefully since it allows to estimate the time for the algorithm to complete and understand if the algorithm got
stuck.
After the installation, it is possible to compute the rwc score directly in the python interpreter by using the function
get_rwc inside the random_walk_controversy package.
>>> from random_walk_controversy import get_rwc
>>> graph = read_edgelist()
>>> side1_nodes = read_nodelist() # list of nodes belonging to partition 1
>>> side2_nodes = read_nodelist() # list of nodes belonging to partition 2
>>> node_percentage = 0.3
>>> number_simulations = 1000
>>> get_rwc(graph, side1_nodes, side2_nodes, node_percentage, number_simulations)
76.233