pdiffsim is a simulation framework for studying linguistic diffusion. It is implemented in Python and makes optional use of the graph-tool library.
Features:
- implemented in Python3
- can be run either locally with graphical output on the screen, or remotely on a server
- two modes: spatial arrangement of agents on a 2D grid, or use of a social network (either use a predefined one or create a randomized one)
- fully customizable through config files
- optional graphical output with GTK/graph-tool (both onscreen and offscreen)
- fairly fast even for large networks up to 2000 agents
- uses R scripts to automatically produce graphs after a simulation run
Requirements:
- Python3
- PIL and/or graph-tool (depends itself on GTK for graphical output)
- Scipy/Numpy
- R language (for automated graphs)
- preferrably a UNIX environment with Makefile / Bash
Demos:
This is a demo that shows a simulation run in which an innovation (green) successfully spreads throughout a social network. In this scenario, the innovation is given a variant bias so that it wins out against the conservative variant (red).
Here we use a network of 200 agents and give the agents a discrete production mechanism (binary choice between innovation or old variant). Using a variant bias, we observe a cascade of change.
Here we use a network of 2000 agents and an alternative diffusion mechanism which produces a wave that diminishes in strength as it travels.
Author:
Luzius Thöny, lucius "dot" antonius "ät" gmail "dot" com