|View source on GitHub|
Computes curve (ROC or PR) values for a prespecified number of points.
tf.contrib.metrics.streaming_curve_points( labels=None, predictions=None, weights=None, num_thresholds=200, metrics_collections=None, updates_collections=None, curve='ROC', name=None )
streaming_curve_points function creates four local variables,
that are used to compute the curve values. To discretize the curve, a linearly
spaced set of thresholds is used to compute pairs of recall and precision
For best results,
predictions should be distributed approximately uniformly
in the range [0, 1] and not peaked around 0 or 1.
For estimation of the metric over a stream of data, the function creates an
update_op operation that updates these variables.
None, weights default to 1. Use weights of 0 to mask values.
Tensorwhose shape matches
predictions. Will be cast to
predictions: A floating point
Tensorof arbitrary shape and whose values are in the range
Tensorwhose rank is either 0, or the same rank as
labels, and must be broadcastable to
labels(i.e., all dimensions must be either
1, or the same as the corresponding
num_thresholds: The number of thresholds to use when discretizing the roc curve.
metrics_collections: An optional list of collections that
aucshould be added to.
updates_collections: An optional list of collections that
update_opshould be added to.
curve: Specifies the name of the curve to be computed, 'ROC' [default] or 'PR' for the Precision-Recall-curve.
name: An optional variable_scope name.
Tensorwith shape [num_thresholds, 2] that contains points of the curve.
update_op: An operation that increments the
labelshave mismatched shapes, or if
Noneand its shape doesn't match
predictions, or if either
updates_collectionsare not a list or tuple.
TODO(chizeng): Consider rewriting this method to make use of logic within the precision_recall_at_equal_thresholds method (to improve run time).