import numpy as np

epsilon = 1e-7

def get_loss_from_list(y_pred_list,y_true_list,target):

return np.mean([get_loss(yg,yp,target) for yp,yg in zip(y_pred_list,y_true_list)])

def get_loss(y_true,y_pred,target):

return target.loss_np(y_true,y_pred)

def mae_np(y_true,y_pred):

return np.mean(np.abs(y_pred-y_true))

def mse_np(y_true,y_pred):

return np.mean((y_pred-y_true)**2)

def binary_crossentropy_np(y_true,y_pred):

y_pred = np.clip(y_pred,epsilon,1-epsilon)

return np.mean(- (y_true*np.log(y_pred) + (1-y_true)*np.log(1 - y_pred)))

def hinge_np(y_true,y_pred):

return np.mean(np.maximum(0.0,1 - y_pred*y_true))

def squared_hinge_np(y_true,y_pred):

return np.mean(np.maximum(0.0,1 - y_pred*y_true)**2)