def calculate_loss(self):
x = self.get_input_sequences()
conv_inputs = self.encode(x, features=self.encode_features)
decode_x = tf.concat([x, 1.0 - tf.expand_dims(self.x_is_zero, 2)],
axis=2)
self.initialize_decode_params(decode_x, features=self.decode_features)
y_hat = self.decode(decode_x,
conv_inputs,
features=self.decode_features)
y_hat, p = tf.unstack(y_hat, axis=2, num=2)
y_hat = tf.nn.sigmoid(p) * (y_hat + self.x_mean)
self.loss = sequence_rmse(self.y,
y_hat,
self.y_len,
weights=self.weights)
self.prediction_tensors = {
'preds': tf.nn.relu(y_hat),
'lengths': self.x_len,
'ids': self.y_id,
}
return self.loss
def calculate_loss(self):
x = self.get_input_sequences()
preds = self.calculate_outputs(x)
loss = sequence_rmse(self.next_reorder_size, preds, self.history_length, 100)
return loss
def calculate_loss(self):
x = self.get_input_sequences()
preds = self.calculate_outputs(x)
loss = sequence_rmse(self.next_reorder_size, preds,
self.history_length, 100)
return loss