U Xe'@sddlZddlmZddlZddlZddlmZddlZ ej ddddZ e dd d Z e d d d Z e ddZej ddddZddZdddZddZddZdS)N)warn)check_is_fittedT)ZparallelcCsXtj|jd|ftjd}t|jdD](}||jdd}|d|}|||<q*|S)aA fast computation of knn indices. Parameters ---------- X: array of shape (n_samples, n_features) The input data to compute the k-neighbor indices of. n_neighbors: int The number of nearest neighbors to compute for each sample in ``X``. Returns ------- knn_indices: array of shape (n_samples, n_neighbors) The indices on the ``n_neighbors`` closest points in the dataset. rdtypeZ quicksort)ZkindN)npemptyshapeZint32numbaprangeZargsort)X n_neighborsZ knn_indicesrowvr)lib/python3.8/site-packages/umap/utils.pyfast_knn_indicess   rz i4(i8[:])cCs|dd@d>d@|dd>d@|dAd?A|d<|dd@d >d@|dd >d@|dAd ?A|d<|d d @d >d@|d d>d@|d Ad?A|d <|d|dA|d AS)zA fast (pseudo)-random number generator. Parameters ---------- state: array of int64, shape (3,) The internal state of the rng Returns ------- A (pseudo)-random int32 value rl l ll r)staterrr tau_rand_int(s rz f4(i8[:])cCst|}tt|dS)aA fast (pseudo)-random number generator for floats in the range [0,1] Parameters ---------- state: array of int64, shape (3,) The internal state of the rng Returns ------- A (pseudo)-random float32 in the interval [0, 1] i)rabsfloat)rZintegerrrrtau_randBs r cCs2d}t|jdD]}|||d7}qt|S)zCompute the (standard l2) norm of a vector. Parameters ---------- vec: array of shape (dim,) Returns ------- The l2 norm of vec. grr)rangerrZsqrt)ZvecresultirrrnormSs r$cCs^|j\}}tj||f|jd}t|D]0}t|D] }|||||ff|||f<q6q(|S)aReturn a submatrix given an orginal matrix and the indices to keep. Parameters ---------- dmat: array, shape (n_samples, n_samples) Original matrix. indices_col: array, shape (n_samples, n_neighbors) Indices to keep. Each row consists of the indices of the columns. n_neighbors: int Number of neighbors. Returns ------- submat: array, shape (n_samples, n_neighbors) The corresponding submatrix. r)rrZzerosrr r )ZdmatZ indices_colr Zn_samples_transformZ n_samples_fitZsubmatr#jrrr submatrixes   r&cCsttS)N)timectimerrrrtssr)cCsT|}tjddt|j|jDtd}|||}tj||||dd|dS)aFind the unique elements of a sparse csr matrix. We don't explicitly construct the unique matrix leaving that to the user who may not want to duplicate a massive array in memory. Returns the indices of the input array that give the unique values. Returns the indices of the unique array that reconstructs the input array. Returns the number of times each unique row appears in the input matrix. matrix: a csr matrix return_index = bool, optional If true, return the row indices of 'matrix' return_inverse: bool, optional If true, return the indices of the unique array that can be used to reconstruct 'matrix'. return_counts = bool, optional If true, returns the number of times each unique item appears in 'matrix' The unique matrix can computed via unique_matrix = matrix[index] and the original matrix reconstructed via unique_matrix[inverse] cSsg|]\}}t||qSr)tuple).0xyrrr szcsr_unique..r) return_indexreturn_inverse return_countsr)ZtolilrZasarrayziprowsdataobjectunique)Zmatrixr/r0r1Z lil_matrixr3Z return_valuesrrr csr_uniques" r7cCsTt|d|jr4t|j|jjdddk}nt|jjdddk}|S)aB Returns a boolean vector indicating which vertices are disconnected from the umap graph. These vertices will often be scattered across the space and make it difficult to focus on the main manifold. They can either be filtered and have UMAP re-run or simply filtered from the interactive plotting tool via the subset_points parameter. Use ~disconnected_vertices(model) to only plot the connected points. Parameters ---------- model: a trained UMAP model Returns ------- A boolean vector indicating which points are disconnected graph_r)Zaxisr)rr6rZarrayr8Z_unique_inverse_sumZflatten)ZmodelZvertices_disconnectedrrrdisconnected_verticess   r:cCsLtj|\}}}t|}tj||d}||}tt|rHtd|S)aCalculate the average distance to each points nearest neighbors. Parameters ---------- dist_matrix: a csr_matrix A distance matrix (usually umap_model.graph_) Returns ------- An array with the average distance to each points nearest neighbors )ZweightszwEmbedding contains disconnected vertices which will be ignored.Use umap.utils.disconnected_vertices() to identify them.)scipyZsparsefindrZbincountanyZisnanr)Z dist_matrixZrow_idxZcol_idxvalZcount_non_zero_elemsZsum_non_zero_elemsZaveragesrrraverage_nn_distances  r?)TTT)r'warningsrZnumpyrr Zsklearn.utils.validationrZ scipy.sparser;Znjitrrr r$r&r)r7r:r?rrrrs&          !