import numpy as np
from sklearn.metrics import roc_curve

def compute_inverse_eer(y_true, scores):

    fpr, tpr, thresholds = roc_curve(y_true, scores, pos_label=1)
    fnr = 1 - tpr
    try:
        eer = fpr[np.nanargmin(np.absolute((fnr - fpr)))]
    except ValueError:
        eer = 0.0  # TODO: fix this
    return 1 - eer