keraflow.objectives Namespace Reference

Built-in objectives functions. More...

Functions
def	accuracy
	Accuracy objective. More...
def	square_error
	Squared error loss.
def	absolute_error
	Absolute error loss.
def	absolute_percentage_error
	Absolute percentage error loss.
def	squared_logarithmic_error
	Squared log error loss.
def	hinge
	Squared hinge error loss.
def	squared_hinge
	Squared hinge error loss.
def	binary_crossentropy
	Crossentropy loss for binary labels. More...
def	categorical_crossentropy
	Crossentropy loss for multiple-class classification. More...
def	kullback_leibler_divergence
	Information gain from a predicted probability distribution Q to a true probability distribution P. More...
def	poisson
	Poisson loss.
def	cosine_proximity
	Cosine proximity loss.

Detailed Description

Built-in objectives functions.

All objectives should accept two 2D (or more) tensors y_pred & y_true and return the resulting 2D (or more) tensor containg objective score per sample.

Function Documentation

def keraflow.objectives.accuracy	(		y_pred,
			y_true
	)

Accuracy objective.

Note that this objective is a switch of binary_accuracy and categorical_accuracy.

def keraflow.objectives.binary_crossentropy	(		y_pred,
			y_true
	)

Crossentropy loss for binary labels.

Shape of y_pred and y_true should be (N,1), where N is the sample size.

def keraflow.objectives.categorical_crossentropy	(		y_pred,
			y_true
	)

Crossentropy loss for multiple-class classification.

y_pred and y_true should be binary matrix of shape (N,c), where N is the sample size, c is the number of class.

def keraflow.objectives.kullback_leibler_divergence	(		y_pred,
			y_true
	)

Information gain from a predicted probability distribution Q to a true probability distribution P.

Gives a measure of difference between both distributions.