Scipy logloss

[Eirik Rosvold Larsen]

Hi,

Is there an out-of-the-box implementation of Python's logloss function in Breeze?

//implementation from kaggle:

https://www.kaggle.com/wiki/LogarithmicLoss

import scipy as sp def logloss(act, pred): epsilon = 1e-15 pred = sp.maximum(epsilon, pred) pred = sp.minimum(1-epsilon, pred) ll = sum(act*sp.log(pred) + sp.subtract(1,act)*sp.log(sp.subtract(1,pred))) ll = ll * -1.0/len(act) return ll

Implementation from skkit-learn:

https://github.com/scikit-learn/scikit-learn/blob/317dea8a05b0087f83318f318d112976e90566ff/sklearn/neural_network/_base.py

def log_loss(y_true, y_prob):
	"""Compute Logistic loss for classification.
	Parameters
	----------
	y_true : array-like or label indicator matrix
	Ground truth (correct) labels.
	y_prob : array-like of float, shape = (n_samples, n_classes)
	Predicted probabilities, as returned by a classifier's
	predict_proba method.
	Returns
	-------
	loss : float
	The degree to which the samples are correctly predicted.
	"""
	y_prob = np.clip(y_prob, 1e-10, 1 - 1e-10)
	if y_prob.shape[1] == 1:
	y_prob = np.append(1 - y_prob, y_prob, axis=1)
	if y_true.shape[1] == 1:
	y_true = np.append(1 - y_true, y_true, axis=1)
	return -np.sum(y_true * np.log(y_prob)) / y_prob.shape[0]

Regards, Eirik

[David Hall]

sorry, not built in.