pyoptes

Python framework for optimization of epidemic testing strategies

requirements

The code should work with the following python environment version combinations:

• python=3.7.6, numba=0.48.0, numpy=1.19.2, scipy=1.4.1, networkx=2.6.3
• python=3.7.6, numba=0.54.1, numpy=1.20.3, scipy=1.4.1, networkx=2.6.3
• python=3.7.12, numba=0.54.1, numpy=1.20.3, scipy=1.7.3, networkx=2.6.3
• python=3.9.9, numba=0.54.1, numpy=1.20.3, scipy=1.7.3, networkx=2.6.3

If the older numba=0.46.0 was used instead, an unresolved issue with data types arises.

If one needs tensorflow/keras, one can create a corresponding env as follows:

• conda create -vv -n "myenv" python=3.9.9=h62f1059_0_cpython
• conda activate myenv
• pip install tensorflow
• conda install numba numpy networkx scipy (When trying to install tensorflow via conda instead, conda may crash) This should give tensorflow 2.7.0.

Getting started

• Try the scripts test_target_function.py, test_target_function_2.py, and test_target_function_3.py to see the target function for optimizing the test budget allocation in action, to get an idea about how stochastic ("noisy") the target function evaluations are, and to see how several heuristically intuitive candidate inputs (=budget allocations) perform.
• Try the scripts small_test.py and larger_test.py to understand how the underlying components of the target function can be used: the synthetic generation of transmissions data (pyoptes.networks.transmissions.scale_free) and the simulation of an epidemiological model (pyoptes.epidemiological_models.si_model_on_transmissions).
• Look into the source codes of pyoptes.networks.transmissions.scale_free and pyoptes.epidemiological_models.si_model_on_transmissions how these work in detail.