U !?h0@sbddlZddlmZmZeddddddZeddddd d d ZGd d d eZddZ dS)N)njitprangezNi8(i8[:], i8[:], i8[:], i8[:], i8[:], f4[:, :, :], f4[:], f4[:, :], i8[:], i8)T)cachenogilc CsT|jd} tdtj} d} d} d| }d| | <| d7} |jd| | <| d7} | | | <| d7} | d7} | dkrP| d8} | | }| d8} | | }| d8} | | }||df}||df}t| D]$}|||f||<|||f||<qt|d|D]H}t| D]:}t|||||f||<t|||||f||<qq|||kr~|||<|||<d||<d||<d||<d||<qf||}t|}||d||}|}|d}||krp||kr|||f|kr|d7}q||kr |||f|kr |d8}q||krt| D]2}|||f}|||f|||f<||||f<q||}||||<|||<q|||f|kr|d7}|}t|| t|| |}| }| d7} | }| d7} || | <| d7} || | <| d7} || | <| d7} || | <| d7} || | <| d7} || | <| d7} |||<|||<|||<|||<|||<|||<qf| S)Nrg?)shapenpemptyint64rangeminmaxZargmax)i0_indsi1_inds less_inds more_inds split_dimsbounds split_valuespointsindices min_leaf_size dimensionstack stack_sizen_nodesZ max_leaf_sizei_nodei1Zi0lohidilengths split_dimZ split_valuejtempZ temp_i_nodeZi_splitlessmorer,G/var/www/html/venv/lib/python3.8/site-packages/pymatting/util/kdtree.py _make_trees                  r.zgvoid(i8[:], i8[:], i8[:], i8[:], i8[:], f4[:, :, :], f4[:], f4[:, :], f4[:, :], i8[:, :], f4[:, :], i8))rrparallelc ! Cs@|jd} t|jdD] } || }| | }| | }tdtj}d}d|d<d}|dkr2|d8}||}|| krd}t| D]@}||}|t||d|ft||d|f|}|||7}q|||dkrqT||dkrt||||D]}d}t| D]&}|||||f}|||7}q|}t|dddD]}|||krB|}qB|| krtt|| d|dD]&}||d||<||d||<qz|||<|||<t|d| }qqT||}||}||} || ||kr|||<|d7}|||<|d7}qT|||<|d7}|||<|d7}qT|d7}qdS)Nrrrr r )r rr r rrrr)!rrrrrrrr query_pointsZ out_indicesZ out_distanceskrZi_queryZ query_point distancesrrZ n_neighborsrr distr$pZdpr%ZdistanceddZ insert_posr(r*r+r'r,r,r- _find_knnsd   (   r6c@s"eZdZdZdddZddZdS) KDTreezKDTree implementationc Cs|jtjkst|j\}}d||d|}t|tj|_t|tj|_t|tj|_ t|tj|_ t|tj|_ t|d|ftj|_ t|tj|_ ||_t|tj|_t|j|j|j |j |j |j |j |j|j| |_dS)aConstructs a KDTree for given data points. The implementation currently only supports data type `np.float32`. Parameters ---------- data_points: numpy.ndarray (of type `np.float32`) Dataset with shape :math:`n \times d`, where :math:`n` is the number of data points in the data set and :math:`d` is the dimension of each data point min_leaf_size: int Minimum number of nodes in a leaf, defaults to 8 Example ------- >>> from pymatting import * >>> import numpy as np >>> data_set = np.random.randn(100, 2) >>> tree = KDTree(data_set.astype(np.float32)) rrN)dtyper float32AssertionErrorr r rrrrrrrrcopyshuffled_data_pointsZarangeZastypeshuffled_indicesr.r)self data_pointsrZn_datarZ max_nodesr,r,r-__init__s0  zKDTree.__init__c Cs|jtjkst|jjd|kr>td|jjdd|d|jd}t||ftj}t||ftj}t |j |j |j |j |j|j|j|j|||| |j|}t|}||fS)a6Query the tree Parameters ---------- query_points: numpy.ndarray (of type `np.float32`) Data points for which the next neighbours should be calculated k: int Number of neighbors to find Returns ------- distances: numpy.ndarray Distances to the neighbors indices: numpy.ndarray Indices of the k nearest neighbors in original data array Example ------- >>> from pymatting import * >>> import numpy as np >>> data_set = np.random.randn(100, 2) >>> tree = KDTree(data_set.astype(np.float32)) >>> tree.query(np.array([[0.5,0.5]], dtype=np.float32), k=3) (array([[0.14234178, 0.15879704, 0.26760164]], dtype=float32), array([[29, 21, 20]])) rzNumber of data points (z4) is less than the number of neighbors requested (k=z<). Please provide a larger dataset or reduce the value of k.)r9r r:r;r=r ValueErrorr rr6rrrrrrrr>sqrt)r?r0r1Zn_queryZsquared_distancesrr2r,r,r-querys2   z KDTree.queryN)r8)__name__ __module__ __qualname____doc__rArDr,r,r,r-r7s .r7cCst|}|||S)aFind k nearest neighbors in a data set. The implementation currently only supports data type `np.float32`. Parameters ---------- data_points: numpy.ndarray (of type `np.float32`) Dataset with shape :math:`n \times d`, where :math:`n` is the number of data points in the data set and :math:`d` is the dimension of each data point query_points: numpy.ndarray (of type `np.float32`) Data points for which the next neighbours should be calculated k: int Number of neighbors to find Returns ------- distances: numpy.ndarray Distances to the neighbors indices: numpy.ndarray Indices of the k nearest neighbors in original data array Example ------- >>> from pymatting import * >>> import numpy as np >>> data_set = np.random.randn(100, 2) >>> knn(data_set.astype(np.float32), np.array([[0.5,0.5]], dtype=np.float32), k=2) (array([[0.16233477, 0.25393516]], dtype=float32), array([[25, 17]])) )r7rD)r@r0r1treer,r,r-knnYsrJ) numpyr Znumbarrr.r6objectr7rJr,r,r,r-s   bm