Source code for monte_carlo_analysis.strategies.ClassWiseDistributionStrategy

"""
**Author** : Robin Camarasa

**Institution** : Erasmus Medical Center

**Position** : PhD student

**Contact** : r.camarasa@erasmusmc.nl

**Date** : 2020-10-29

**Project** : monte_carlo_analysis

**Implement ClassWiseDistributionStrategy class**

"""
import numpy as np
from itertools import product
from monte_carlo_analysis.strategies import Strategy
from numba import jit


[docs]class ClassWiseDistributionStrategy(Strategy):
[docs]    @staticmethod
    @jit
    def transformation(
            uncertainty_metric_transformation: callable,
            ensemble_output: np.array, counter: list
            ) -> np.array:
        """
        Compute the strategy

        :param ensemble_output: Ensemble error
        """
        # Get the dim size of the output uncertainty map
        uncertainty_map_shape = ensemble_output.shape[2:]

        # Initialize uncertainty metric
        uncertainty_map = np.zeros(
            (ensemble_output.shape[1], ensemble_output.shape[1]) +\
            uncertainty_map_shape
            )

        # Loop over the indexes of the output uncertainty metric
        for i in counter:
            for j1 in range(ensemble_output.shape[1]):
                for j2 in range(j1, ensemble_output.shape[1]):
                    uncertainty_map[(j1, j2,) + i] = uncertainty_metric_transformation(
                        ensemble_output[(slice(None), j1,) + i],
                        ensemble_output[(slice(None), j2,) + i]
                        )
                    uncertainty_map[j2, j1, :] = uncertainty_map[j1, j2, :]
        return np.max(uncertainty_map, axis=0)