Getting started with EpiStochModels

EpistochModels can be installed from source (if you clone our Github repository).

But first you must have a D compiler installed. If your operating system does not have a pre-built package:

$ curl https://dlang.org/install.sh | bash -s

Then you can build and install the Python package.

$ python setup.py build
$ python setup.py install

We will publish the package on PyPI soon so that you will be able to simply pip install it.

Using the Built-in Models

The models can be used from Python (see notebooks) or straight from D code:

/+dub.json:
{
	"name": "sir_example",
	"authors": [
		"Flávio Codeço Coelho"
	],
	"description": "Example usage of SIR model from epistochmodels",
	"copyright": "Copyright © 2020, Flávio Codeço Coelho",
	"license": "GPL-3",
    "dependencies": {
		"epistochmodels": "*"
	}
}
+/
import std.datetime.stopwatch;
import std.stdio;
import models;

void main(){
    double beta = 0.7;
    double gam = 0.3;
    int N = 1_000_000;
    int I0 = 10;
    double tf = 1000;
    auto sw = StopWatch(AutoStart.no);
    auto model = new SIR(N, beta, gam);
    model.initialize(N-I0, I0, 0);
    sw.start();
    auto sim = model.run(0, tf);
    sw.stop();
    writeln("Time of the SIR run with N=1000000: ", sw.peek());
    writefln("Number of steps: %s", sim[0].length);
}

The D code above can be run with the following command:

$ dub run --build=release --single sir_example.d

Implementing your own CTMC models

If you are willing to write D code, EpiStochModels also provides a class to simulate any CTMC model you want, as long as you provide it with a transition matrix and propensity functions.

The code below is the implementation of a SIRD model.

int[][] tmat = [[-1,1,0,0],
                [0,-1,1,0],
                [0,-1,0,1]];
double[string] pars = [
    "beta": 0.3,
    "gam": 0.05,
    "mu": 0.01
];
alias double delegate(int[], double[string]) dy;
dy[] props = [
    (v,p)=> p["beta"]*v[0]*v[1]/sum(v[0..$-1]), // infection
    (v,p)=> p["gam"]*v[1], // recovery
    (v,p)=> p["mu"]*v[1] // death
    ];
CTMC model = new CTMC(tmat, props);
model.initialize([995,5,0,0],pars);
auto res = model.run(0, 1000);