sassonlan / dynamics_sim Goto Github PK

A generic, extensible simulation library for evolutionary game theory simulations. DyPy provides Moran, and Wright-Fisher processes, as well as Replicator Dynamics and makes it simple to execute complex and robust simulations across a range of parameters and visualize the results with beautiful graphs.

See documentation here.

####Requirements

DyPy depends on matplotlib for graphing, and numpy and joblib. To install these dependencies, make sure you are in the root directory of the repo and run the following command, which may require sudo.

$ pip install -r requirements.txt

####Usage

The easiest way to get started with DyPy is to subclass the Game class and define the game that of interest to be simulated by defining its payoff matrix appropriately as a function of various parameters. You can also define a function that classifies equlibria as a function of the distribution of players playing each strategy.

Once the game class is defined, choose a dynamics process and execute the desired simulation. Some options are:

• Simulate a given number of generations of one simulation, and graph the dynamics of each player's strategies over time
• Repeat a given simulation multiple times and return the frequency of each resulting equilibria.
• Vary one or more parameters to the dynamics or game constructors and graph the effect of this variation on the resulting equilibria, either in 2D or 2D graphs.

The GameDynamicsWrapper and VariedGame classes take care of simplifying the simulation and graphing processes, and automatically parallelize the computations across all available cores.

To see an example, take a look at the Cooperate Without Looking subclass along with its associated simulations.

####Persistence (coming soon)

DyPy decouples the process of simulating with graphing. This encourages users to run long-running simulations and gather tons of data, and then insert and tweak the graph parameters afterwards.