deepmie/utils.py

# this file collects several utility functions
import numpy as np
from keras import backend as K

def naive_percent_loss(y_true, y_pred):
    """TODO: Docstring for .
    # Sums up the relative errors at all wavelength points of a spectrum
    # in comparison to a target spectrum
    : y_true -- target spectrum
    : y_pred -- predicted spectrum
    Note that the user needs to ensure we are comparing similiar spectra
    """
    y_err = np.abs(y_true - y_pred) / y_true
    y_err_f = K.flatten(y_err)
    return K.sum(y_err_f)

def calc_mre(y_true, y_pred):
    """TODO: Docstring for .
    # Gives the mean relative errors in percentage terms comparing two sets of spectra
    """
    y_err = np.abs(y_true - y_pred) / y_true
    y_err = 100*np.abs(y_true - y_pred)/y_true
    return np.mean(y_err)

#legacy function to compare two size vectors, not really needed anymore
# def size_percent_loss(y_true, y_pred):
#     y_true_a = 0.5*y_true*(size_max - size_min) + size_av
#     y_pred_a = 0.5*y_pred*(size_max - size_min) + size_av
#     y_err = np.abs(y_true_a - y_pred_a)/y_true_a
#     y_err_f = K.flatten(y_err)
#     return K.sum(y_err_f)