def padded_accuracy(predictions,
labels,
weights_fn=common_layers.weights_nonzero):
"""Percentage of times that predictions matches labels on non-0s."""
with tf.variable_scope("padded_accuracy", values=[predictions, labels]):
padded_predictions, padded_labels = common_layers.pad_with_zeros(
predictions, labels)
weights = weights_fn(padded_labels)
outputs = tf.to_int32(tf.argmax(padded_predictions, axis=-1))
padded_labels = tf.to_int32(padded_labels)
return tf.to_float(tf.equal(outputs, padded_labels)), weights
def padded_log_poisson(predictions,
labels,
weights_fn=common_layers.weights_all):
# Expects predictions to already be transformed into log space
predictions, labels = common_layers.pad_with_zeros(predictions, labels)
targets = labels
weights = weights_fn(targets)
lp_loss = tf.nn.log_poisson_loss(targets,
predictions,
compute_full_loss=True)
return tf.reduce_sum(lp_loss * weights), tf.reduce_sum(weights)
def padded_variance_explained(predictions,
labels,
weights_fn=common_layers.weights_all):
# aka R^2
predictions, labels = common_layers.pad_with_zeros(predictions, labels)
targets = labels
weights = weights_fn(targets)
y_bar = tf.reduce_mean(weights * targets)
tot_ss = tf.reduce_sum(weights * tf.pow(targets - y_bar, 2))
res_ss = tf.reduce_sum(weights * tf.pow(targets - predictions, 2))
r2 = 1. - res_ss / tot_ss
return r2, tf.reduce_sum(weights)
def padded_sequence_accuracy(predictions,
labels,
weights_fn=common_layers.weights_nonzero):
"""Percentage of times that predictions matches labels everywhere (non-0)."""
with tf.variable_scope("padded_sequence_accuracy",
values=[predictions, labels]):
padded_predictions, padded_labels = common_layers.pad_with_zeros(
predictions, labels)
weights = weights_fn(padded_labels)
outputs = tf.to_int32(tf.argmax(padded_predictions, axis=-1))
padded_labels = tf.to_int32(padded_labels)
not_correct = tf.to_float(tf.not_equal(outputs,
padded_labels)) * weights
axis = list(range(1, len(outputs.get_shape())))
correct_seq = 1.0 - tf.minimum(1.0,
tf.reduce_sum(not_correct, axis=axis))
return correct_seq, tf.constant(1.0)
def padded_accuracy_topk(predictions,
labels,
k,
weights_fn=common_layers.weights_nonzero):
"""Percentage of times that top-k predictions matches labels on non-0s."""
with tf.variable_scope("padded_accuracy_topk",
values=[predictions, labels]):
padded_predictions, padded_labels = common_layers.pad_with_zeros(
predictions, labels)
weights = weights_fn(padded_labels)
effective_k = tf.minimum(k, tf.shape(padded_predictions)[-1])
_, outputs = tf.nn.top_k(padded_predictions, k=effective_k)
outputs = tf.to_int32(outputs)
padded_labels = tf.to_int32(padded_labels)
padded_labels = tf.expand_dims(padded_labels, axis=-1)
padded_labels += tf.zeros_like(outputs) # Pad to same shape.
same = tf.to_float(tf.equal(outputs, padded_labels))
same_topk = tf.reduce_sum(same, axis=-1)
return same_topk, weights
def padded_rmse(predictions, labels, weights_fn=common_layers.weights_all):
predictions, labels = common_layers.pad_with_zeros(predictions, labels)
targets = labels
weights = weights_fn(targets)
error = tf.sqrt(tf.pow(predictions - labels, 2))
return tf.reduce_sum(error * weights), tf.reduce_sum(weights)