Epidemics Simulation

Simulation of disease epidemics, on small world networks.

The simulation accounts for different infection rates, different synergy parameters (collaborative or competitive diseases), and different network sizes.

Installation

• Download this repository
• Install a virtual python environment (preferred python version is 3.7):

python -m venv venv
source venv/bin/activate

• Install dependencies using pip install -r requirements.txt

Running a single simulation

Simulation are run using the Network classes.

For instance, a rectangular network:

from src.world.network.rectilinear import Rectilinear2DNetwork

my_simulation = Rectilinear2DNetwork(shape=(20, 20))
my_simulation.start_infection_at_center()
my_simulation.run()

This runs a single simulation, but does not output or save any results.

Creating videos

A video of the simulation can be created by using the SnapshotHistoryCallback class:

callback = SnapshotHistoryCallback()

my_simulation = Rectilinear2DNetwork(shape=(20, 20))
my_simulation.start_infection_at_center()
my_simulation.run(callback=callback)

callback.make_video(filename="simulation.mp4")

This will run a simulation and save a video visualation of the map over the duration of the simulation.

Using MLFlow to save results

Simulations cab be tracked using the MLFlowCallback class:

from src.world.network.rectilinear import Rectilinear2DNetwork

callback = MLFlowCallback(log_histories=False)

my_simulation = Rectilinear2DNetwork(shape=(20, 20))
my_simulation.start_infection_at_center()
my_simulation.run(callback=callback)

This will save the parameters and results of the simulation. Parameters are input choices, such as alpha (the synergy-free infection rate), beta (the synergy parameter), and network size. Output results include the percentage of cells infected at end; the final epidemic state of the simulation; etc.

The MLFlowCallback class has optional flag log_histories, which can be used to save the output parameters at all time steps in the simulation.

Viewing results from MFLow

MFFlow results can be viewed using mlflow ui and opening your browser to the http:// address specified in the terminal output. The MLFlow UI displays each simulations as a single row, with input parameters and output parameters displayed as columns.

Running many simulations

Large sets of simulations are most easily run using the script mlflow_scripts/run_fill.py. This script runs a large number of simulations, defined according to the config file (default config.yml).

The default config.yml file runs at most 100 simulations (maximum_runs) with ranges of parameter values (metrics.ALPHA etc). The simulations are run in batches of size 50 (runs_per_batch) and saved to MLFlow database under the experiment name synergy.

The script mlflow_scripts/visualise.py can also be used to visualise how many simulations have been run for each set of parameters.