Skip to content

Latest commit

 

History

History
297 lines (224 loc) · 6.66 KB

ClassificationMatch.md

File metadata and controls

297 lines (224 loc) · 6.66 KB

TFSimilarity.callbacks.ClassificationMatch

Abstract base class for defining the classification matching strategy.

Inherits From: ABC

TFSimilarity.callbacks.ClassificationMatch(
    name: str = , canonical_name: str = 
) -> None

Attributes

name Name associated with the match object, e.g., match_nearest
canonical_name The canonical name associated with match strategy, e.g., match_nearest
distance_thresholds The max distance below which a nearest neighbor is considered a valid match. Defaults to None and must be set using ClassificationMatch.compile()``` .
count The total number of queries.
fn The count of False negatives matches.

A False negative match is when the query label == the label generated by a matcher and the distance to the query is greater than the distance threshold.
fp The count of False positive matches.

A False positive match is when the query label != the label generated by a matcher and the distance to the query is less than the distance threshold.
tn The count of True negatives matches.

A True negative match is when the query label != the label generated by a matcher and the distance to the query is greater than the distance threshold.
tp The count of True positive matches.

A True positive match is when the query label == the label generated by a matcher and the distance to the query is less than the distance threshold.

Methods

compile

View source

compile(
    distance_thresholds: Optional[<a href="../../TFSimilarity/callbacks/FloatTensor.md">TFSimilarity.callbacks.FloatTensor```
</a>] = None
)

Configures the distance thresholds used during matching.

Args
distance_thresholds The max distance below which a nearest neighbor is considered a valid match. A threshold of math.inf is used if None is passed.

compute_count

View source

compute_count(
    query_labels: <a href="../../TFSimilarity/callbacks/IntTensor.md">TFSimilarity.callbacks.IntTensor```
</a>,
    lookup_labels: <a href="../../TFSimilarity/callbacks/IntTensor.md">TFSimilarity.callbacks.IntTensor```
</a>,
    lookup_distances: <a href="../../TFSimilarity/callbacks/FloatTensor.md">TFSimilarity.callbacks.FloatTensor```
</a>
) -> None

Computes the match counts at each of the distance thresholds.

This method computes the following at each distance threshold.

  • True Positive: The query label matches the derived lookup label and the derived lookup distance is <= the current distance threshold.

  • False Positive: The query label does not match the derived lookup label but the derived lookup distance is <= the current distance threshold.

  • False Negative: The query label matches the derived lookup label but the derived lookup distance is > the current distance threshold.

  • True Negative: The query label does not match the derived lookup label and the derived lookup distance is > the current distance threshold.

Note: compile must be called before calling match.

Args
query_labels A 1D array of the labels associated with the queries.
lookup_labels A 2D array where the jth row is the labels associated with the set of k neighbors for the jth query.
lookup_distances A 2D array where the jth row is the distances between the jth query and the set of k neighbors.

derive_match

View source

derive_match(
    lookup_labels: <a href="../../TFSimilarity/callbacks/IntTensor.md">TFSimilarity.callbacks.IntTensor```
</a>,
    lookup_distances: <a href="../../TFSimilarity/callbacks/FloatTensor.md">TFSimilarity.callbacks.FloatTensor```
</a>
) -> Tuple[<a href="../../TFSimilarity/callbacks/IntTensor.md">TFSimilarity.callbacks.IntTensor``<b>
</a>, <a href="../../TFSimilarity/callbacks/FloatTensor.md">TFSimilarity.callbacks.FloatTensor</b>``
</a>]

Derive a match label and distance from a set of K neighbors.

For each query, derive a single match label and distance given the associated set of lookup labels and distances.

Args
lookup_labels A 2D array where the jth row is the labels associated with the set of k neighbors for the jth query.
lookup_distances A 2D array where the jth row is the distances between the jth query and the set of k neighbors.
Returns
A Tuple of FloatTensors: derived_labels: A FloatTensor of shape - [len(lookup_labels), 1] where the jth row contains the derived label for the jth query.

derived_distances: A FloatTensor of shape

  • [len(lookup_labels), 1] where the jth row contains the distance associated with the jth derived label.

get_config

View source

get_config()