Source code for pyreco.metrics

import numpy as np
from sklearn.metrics import r2_score




def mse(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    # calculate the difference between true and predicted values
    err = y_true - y_pred

    # square the difference
    squared_err = np.square(err)

    # calculate the mean of the squared differences
    return np.mean(squared_err)





def mae(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    # calculate the difference between true and predicted values
    err = y_true - y_pred

    # calculate the absolute difference between true and predicted values
    absolute_err = np.abs(err)

    # calculate the mean of the absolute differences
    return np.mean(absolute_err)





def r2(y_true: np.ndarray, y_pred: np.ndarray) -> float:
    # computes R2 score for scalar regression problems
    # we will flatten along time and state dimensions
    y_true = y_true.flatten()
    y_pred = y_pred.flatten()

    return r2_score(y_true=y_true, y_pred=y_pred)





def available_metrics() -> list:
    return ["mse", "mean_squared_error", "mae", "mean_absolute_error", "r2", "r2_score"]





def assign_metric(metric: str):
    if metric not in available_metrics():
        raise (
            ValueError(
                f"metric {metric} is not implemented. Check for implementation in metrics.py and in "
                f"assign_metric() function"
            )
        )

    if metric == "mse" or metric == "mean_squared_error":
        return mse
    elif metric == "mae" or metric == "mean_absolute_error":
        return mae
    elif metric == "r2" or metric == "r2_score":
        return r2