compartmental

Utility tools for Approximate Bayesian computation on compartmental models

SIR Example

(Example taken from examples)

To make a SIR model you will need a configuration and an evolution function:

sir_model = {
    "simulation": {
        "n_simulations": 100000,
        "n_executions": 1,
        "n_steps": 130
    },
    "compartments": {
        "S": { 
            "initial_value": 1,
            "minus_compartments": "I"
        },
        "I": { 
            "initial_value": "Io",
        },
        "R": { "initial_value": 0 },
    },
    "params": {
        "betta": {
            "min": 0.1,
            "max": 0.4
        },
        "mu": {
            "min": 0.01,
            "max": 0.2
        },
        "Io": {
            "min": 1e-6,
            "max": 1e-4
        }
    },
    "fixed_params": {
        "K_mean": 1
    },
    "reference": {
        "compartments" : ["R"]
    },
    "results": {
        "save_percentage": 0.1
    }
}

import compartmental as gcm
gcm.use_cupy() // or numpy in case you don't have access to a gpu.


SirModel = gcm.GenericModel(sir_model)

def evolve(m, *args, **kargs):
    p_infected = m.betta * m.K_mean * m.I

    m.R += m.mu * m.I
    m.I += m.S * p_infected - m.I * m.mu
    m.S -= m.S * p_infected

SirModel.evolve = evolve

That's it!

You can now execute it on your GPU and fit the model to some data. Have a look at the example!

Or have a look into a more elavorated model here.

Instalation

compartmental releases are available as wheel packages on PyPI. You can install the last version using pip:

pip install compartmental

Documentation

Documentations is automatically generated from code on main push and hosted in github-pages here.

Help

Just open an issue with the question tag (or clic here), I would love to help!

Contributing

You can contribute with:

• Examples
• Documentation
• Bug report/fix
• Features
• Code

Even only feedback is greatly apreciated.

Just create an issue and let me know you want to help!

Licensing

compartmental is released under the Apache License Version 2.0.