The folder ./src contain all methods code with version serial and parallels. The version in parallel is for run just stochastic models, where you need to define number of replicates for calibrate the mean of simulations.

The folders ./Hybrid and ./Verhulst contain scripts (hybrid.py and verhulst.py) to generate the AGPR and executing the methods ABC MCMC and ABC SMC with and without AGPR. The ABC methods will save sample parameters that generate the posterior distribution in .dat files inside the ./Calibration folder. The AGPR will save a file Dictionary.pkl in ./AGPR composed by {"GPR","samples_parameters","difference_std"}.

• "GPR" - model trained to generate predictions
• "samples_parameters" - samples of parameters used for training
• "difference_std" - difference between max(std) and mean(std) of elements in parametric hypercube.

In case of the hybrid model, the script hybrid.py needs to be executed in parallel using more than one task. The number of tasks must be equal to the number of replicates + 1 of the hybrid model. For example, if you want to define 46 replicates to run the hybrid model, you can define 47 tasks (mpiexec -n 47 python hybrid.py).

To visualize the results you can use the plot.py scripts.

Required packages: (versions used)

• NumPy (version: 1.20.3)
• scikit-learn (version: 1.0.2)
• pickle (version: 4.0)
• seaborn (version: 0.10.0)
• matplotlib (version: 3.2.0)
• scipy (version: 1.7.1)
• mpi4py (version: 3.0.3)