strlearn/utils/base.py
import numpy as np
def scores_to_cummean(scores):
"""
Convert evaluator scores to accumulative mean.
It's the best way to make reader capable to understand anything from your results.
:type scores: array-like, shape (n_estimators, n_chunks, n_metrics)
:param scores: Evaluation scores.
:rtype: array-like, shape (n_estimators, n_chunks, n_metrics)
:returns: Evaluation scores in format possible to read for human being.
.. image:: plots/cummean.png
"""
divider = np.arange(1,scores.shape[1]+1)
cs = np.cumsum(scores, axis=1)
return cs / divider[np.newaxis, :, np.newaxis]