napf.base.KDT#

class napf.base.KDT(tree_data, metric=2, leaf_size=10, nthread=1)[source]#

Bases: object

napf is implemented as template, thus, there are separate classes for each {data_type, dim, metric}. Currently following combinations are supported: data_type: {double, int} dim: {1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20} metric: {L1, L2}

Given tree_data, creates corresponding core kdt class. Tree is initialized using newtree().

Parameters:

tree_data ((n, dim) np.ndarray) – Default is None. {double, float, int, long}
metric (int or str) – Default is 2 and distance will be a squared euklidian distance. Valid options are {1, l1, L1, 2, l2, L2}.
leaf_size (int)
nthread (int) – Default thread count for all multi-thread-

Returns:

core_obj

Return type:

KDT{data_t}D{dim}L{metric}

Methods

`KDT.knn_search`(queries, kneighbors[, nthread])	k-nearest-neighbor search.
`KDT.newtree`(tree_data[, metric, leaf_size, ...])	Given 2D array-like tree_data, it:
`KDT.query`(queries[, nthread])	scipy-like KDTree query call.
`KDT.query_ball_point`(queries, radius, ...[, ...])	scipy-like KDTree query_ball_point call.
`KDT.radii_search`(queries, radii, return_sorted)	Similar to radius_search, but you can specify radius for each query.
`KDT.radius_search`(queries, radius, return_sorted)	Searches for neighbors in given radius.
`KDT.rknn_search`(queries, radius, n_nearest)	Searches for k-nearest neighbors within the radius.
`KDT.unique_data_and_inverse`(radius[, ...])	Finds unique tree data with in given radius tolerance.

Attributes

`KDT.core_tree`	Returns initialized core tree, if there's any.
`KDT.dtype`	Returns dtype of current tree
`KDT.nthread`	Returns saved default value for nthread.
`KDT.tree_data`	Returns data used to initialize core tree.