Functions
def	masked_loss (y_true, y_pred)

def	masked_loss_binary (y_true, y_pred)

def	masked_loss_categorical (y_true, y_pred)

def	loss_binary_crossentropy (y_true, y_pred)

def	loss_categorical_crossentropy (y_true, y_pred)

def	multitask_loss (y_true, y_pred)

def	multitask_loss_weighted (y_true, y_pred)

Function Documentation

def my_losses.loss_binary_crossentropy	(	y_true,
		y_pred
	)

Definition at line 22 of file my_losses.py.

 def loss_binary_crossentropy(y_true, y_pred):
     # Avoid divide by 0
     y_pred = K.clip(y_pred, K.epsilon(), 1 - K.epsilon())
 
     return losses.binary_crossentropy(y_true, y_pred)
 

def my_losses.loss_categorical_crossentropy	(	y_true,
		y_pred
	)

Definition at line 28 of file my_losses.py.

 def loss_categorical_crossentropy(y_true, y_pred):
     # Avoid divide by 0
     y_pred = K.clip(y_pred, K.epsilon(), 1 - K.epsilon())
 
     return losses.categorical_crossentropy(y_true, y_pred)

def my_losses.masked_loss	(	y_true,
		y_pred
	)

Definition at line 4 of file my_losses.py.

 def masked_loss(y_true, y_pred):
     mask_value = -1
     mask = K.cast(K.not_equal(y_true, mask_value), K.floatx())
         
     return multitask_loss(y_true * mask, y_pred * mask)
 

def my_losses.masked_loss_binary	(	y_true,
		y_pred
	)

Definition at line 10 of file my_losses.py.

 def masked_loss_binary(y_true, y_pred):
     mask_value = -1
     mask = K.cast(K.not_equal(y_true, mask_value), K.floatx())
         
     return loss_binary_crossentropy(y_true * mask, y_pred * mask)
 

def my_losses.masked_loss_categorical	(	y_true,
		y_pred
	)

Definition at line 16 of file my_losses.py.

 def masked_loss_categorical(y_true, y_pred):
     mask_value = -1
     mask = K.cast(K.not_equal(y_true, mask_value), K.floatx())
         
     return loss_categorical_crossentropy(y_true * mask, y_pred * mask)
 

def my_losses.multitask_loss	(	y_true,
		y_pred
	)

Definition at line 35 of file my_losses.py.

 def multitask_loss(y_true, y_pred):
     # Avoid divide by 0
     y_pred = K.clip(y_pred, K.epsilon(), 1 - K.epsilon())
 
     # Multi-task loss
     return K.mean(K.sum(- y_true * K.log(y_pred) - (1 - y_true) * K.log(1 - y_pred), axis=1))
 

def my_losses.multitask_loss_weighted	(	y_true,
		y_pred
	)

Definition at line 42 of file my_losses.py.

 def multitask_loss_weighted(y_true, y_pred):
     # Avoid divide by 0
     y_pred = K.clip(y_pred, K.epsilon(), 1 - K.epsilon())
 
     print "...WEIGHTS..."
     #weights = K.cast([1.5, 1.5, 1.5, 1.5, 0.75, 0.75, 0.75, 0.75], K.floatx())
     #weights = K.cast([1.75, 1.75, 1.75, 1.75, 0.5, 0.5, 0.5, 0.5], K.floatx())
     weights = K.cast([5.0*0.67049114189459802, 5.0*0.6574572346004135, 5.0*2.4884334247030484,  5.0*1.7074017025820696, 
                       0.5*1.1687034551437478,  0.5*0.8630824121263583, 0.5*0.6372003918709082,  0.5*2.4018367377204779], K.floatx())
 
     # Multi-task loss
     return K.mean(K.sum((- y_true * K.log(y_pred) - (1 - y_true) * K.log(1 - y_pred)) * weights, axis=1))
 
 

Functions

Function Documentation