B \h=@sddlmZddlZddlmZmZ m Z ddZ ddZ dd Zd d Zd d ZddZddZddZddZddZddZddZddZddZd d!Zd"d#Zd$d%Zd1d'd(Zd2d)d*Zd+d,Zd-d.Z d/d0Z!dS)3N)csr_mean_variance_axis0csc_mean_variance_axis0incr_mean_variance_axis0cCs,t|r|jnt|}d|}t|dS)z2Raises a TypeError if X is not a CSR or CSC matrixz,Expected a CSR or CSC sparse matrix, got %s.N)spZissparseformattype TypeError)XZ input_typeerrr 8lib/python3.7/site-packages/sklearn/utils/sparsefuncs.py_raise_typeerrorsrcCs|dkrtd|dS)N)rrz8Unknown axis value: %d. Use 0 for rows, or 1 for columns) ValueError)axisr r r _raise_error_wrong_axissrcCs6|jd|jdkst|j|j|jdd9_dS)aInplace column scaling of a CSR matrix. Scale each feature of the data matrix by multiplying with specific scale provided by the caller assuming a (n_samples, n_features) shape. Parameters ---------- X : CSR matrix with shape (n_samples, n_features) Matrix to normalize using the variance of the features. scale : float array with shape (n_features,) Array of precomputed feature-wise values to use for scaling. rrZclip)modeN)shapeAssertionErrordataZtakeindices)r scaler r r inplace_csr_column_scalesrcCs:|jd|jdkst|jt|t|j9_dS)a Inplace row scaling of a CSR matrix. Scale each sample of the data matrix by multiplying with specific scale provided by the caller assuming a (n_samples, n_features) shape. Parameters ---------- X : CSR sparse matrix, shape (n_samples, n_features) Matrix to be scaled. scale : float array with shape (n_samples,) Array of precomputed sample-wise values to use for scaling. rN)rrrnprepeatdiffindptr)r rr r r inplace_csr_row_scale.srcCsdt|t|tjr0|dkr$t|St|jSn0t|tjrX|dkrLt|St|jSnt|dS)aCompute mean and variance along an axix on a CSR or CSC matrix Parameters ---------- X : CSR or CSC sparse matrix, shape (n_samples, n_features) Input data. axis : int (either 0 or 1) Axis along which the axis should be computed. Returns ------- means : float array with shape (n_features,) Feature-wise means variances : float array with shape (n_features,) Feature-wise variances rN) r isinstancer csr_matrix_csr_mean_var_axis0_csc_mean_var_axis0T csc_matrixr)r rr r r mean_variance_axis@s    r$cCst|t|tjr@|dkr,t||||dSt|j|||dSn@t|tjrx|dkrdt||||dSt|j|||dSnt|dS)aCompute incremental mean and variance along an axix on a CSR or CSC matrix. last_mean, last_var are the statistics computed at the last step by this function. Both must be initialized to 0-arrays of the proper size, i.e. the number of features in X. last_n is the number of samples encountered until now. Parameters ---------- X : CSR or CSC sparse matrix, shape (n_samples, n_features) Input data. axis : int (either 0 or 1) Axis along which the axis should be computed. last_mean : float array with shape (n_features,) Array of feature-wise means to update with the new data X. last_var : float array with shape (n_features,) Array of feature-wise var to update with the new data X. last_n : int with shape (n_features,) Number of samples seen so far, excluded X. Returns ------- means : float array with shape (n_features,) Updated feature-wise means. variances : float array with shape (n_features,) Updated feature-wise variances. n : int with shape (n_features,) Updated number of seen samples. Notes ----- NaNs are ignored in the algorithm. r) last_meanlast_varlast_nN)rrrr_incr_mean_var_axis0r"r#r)r rr%r&r'r r r incr_mean_variance_axises+      r)cCs>t|tjrt|j|n t|tjr2t||nt|dS)aInplace column scaling of a CSC/CSR matrix. Scale each feature of the data matrix by multiplying with specific scale provided by the caller assuming a (n_samples, n_features) shape. Parameters ---------- X : CSC or CSR matrix with shape (n_samples, n_features) Matrix to normalize using the variance of the features. scale : float array with shape (n_features,) Array of precomputed feature-wise values to use for scaling. N)rrr#rr"rrr)r rr r r inplace_column_scales    r*cCs>t|tjrt|j|n t|tjr2t||nt|dS)a Inplace row scaling of a CSR or CSC matrix. Scale each row of the data matrix by multiplying with specific scale provided by the caller assuming a (n_samples, n_features) shape. Parameters ---------- X : CSR or CSC sparse matrix, shape (n_samples, n_features) Matrix to be scaled. scale : float array with shape (n_features,) Array of precomputed sample-wise values to use for scaling. N)rrr#rr"rrr)r rr r r inplace_row_scales    r+cCszx$||gD]}t|tjr tdq W|dkr<||jd7}|dkrR||jd7}|j|k}||j|j|k<||j|<dS)a5 Swaps two rows of a CSC matrix in-place. Parameters ---------- X : scipy.sparse.csc_matrix, shape=(n_samples, n_features) Matrix whose two rows are to be swapped. m : int Index of the row of X to be swapped. n : int Index of the row of X to be swapped. z m and n should be valid integersrN)rrndarrayr rr)r mntZm_maskr r r inplace_swap_row_cscs   r0c Cs|x$||gD]}t|tjr tdq W|dkr<||jd7}|dkrR||jd7}||krd||}}|j}||}||d}||}||d}||} ||} | | kr|j|d|| | 7<|| |j|d<|| |j|<t|jd||j|||j|||j|||j|dg|_t|jd||j|||j|||j|||j|dg|_dS)a5 Swaps two rows of a CSR matrix in-place. Parameters ---------- X : scipy.sparse.csr_matrix, shape=(n_samples, n_features) Matrix whose two rows are to be swapped. m : int Index of the row of X to be swapped. n : int Index of the row of X to be swapped. z m and n should be valid integersrrN) rrr,r rrZ concatenaterr) r r-r.r/rZm_startZm_stopZn_startZn_stopZnz_mZnz_nr r r inplace_swap_row_csrs<           r2cCs@t|tjrt|||n"t|tjr4t|||nt|dS)a: Swaps two rows of a CSC/CSR matrix in-place. Parameters ---------- X : CSR or CSC sparse matrix, shape=(n_samples, n_features) Matrix whose two rows are to be swapped. m : int Index of the row of X to be swapped. n : int Index of the row of X to be swapped. N)rrr#r0rr2r)r r-r.r r r inplace_swap_row$s   r3cCsl|dkr||jd7}|dkr,||jd7}t|tjrFt|||n"t|tjr`t|||nt|dS)aF Swaps two columns of a CSC/CSR matrix in-place. Parameters ---------- X : CSR or CSC sparse matrix, shape=(n_samples, n_features) Matrix whose two columns are to be swapped. m : int Index of the column of X to be swapped. n : int Index of the column of X to be swapped. rrN)rrrr#r2rr0r)r r-r.r r r inplace_swap_column;s  r4cCs.tt|j}||j|j|}||fS)N)rZ flatnonzerorrZreduceatr)r Zufunc major_indexvaluer r r _minor_reduceVsr7c Cs|j|}|dkrtd|jd|}|dkr8|n|}|t||\}}t|j||k}|||d||<|dk} t | |}t | |}|dkrt j |t t ||ff|jd|fd} n(t j ||t t |ff|j|dfd} | jS)Nrz&zero-size array to reduction operationr)dtyper)rrZtocscZtocsrZsum_duplicatesr7rrrcompressrZ coo_matrixzeroslenr8Aravel) r r min_or_maxNMZmatr5r6Znot_fullmaskZresr r r _min_or_max_axis\s$   rBcCs|dkrdd|jkrtd|jd}|jdkr4|S||j}|jt |jkr`|||}|S|dkrt|d7}|dks|dkrt |||StddS)Nrz&zero-size array to reduction operationr1rz.invalid axis, use 0 for rows, or 1 for columns) rrr8rnnzreducerr=rproductrB)r rr>Zzeror-r r r _sparse_min_or_maxss     rFcCst||tjt||tjfS)N)rFrZminimumZmaximum)r rr r r _sparse_min_maxs rGcCst||tjt||tjfS)N)rFrZfminZfmax)r rr r r _sparse_nan_min_maxs rHFcCsBt|tjst|tjr6|r(t||dSt||dSnt|dS)a^Compute minimum and maximum along an axis on a CSR or CSC matrix and optionally ignore NaN values. Parameters ---------- X : CSR or CSC sparse matrix, shape (n_samples, n_features) Input data. axis : int (either 0 or 1) Axis along which the axis should be computed. ignore_nan : bool, default is False Ignore or passing through NaN values. .. versionadded:: 0.20 Returns ------- mins : float array with shape (n_features,) Feature-wise minima maxs : float array with shape (n_features,) Feature-wise maxima )rN)rrrr#rHrGr)r rZ ignore_nanr r r min_max_axiss  rIcCs|dkrd}n(|dkrd}n|jdkr6td|j|dkrb|dkrL|jStt|j|Sn|dkrt|j}|dkr|S||S|dkr|dkrtj|j|j ddSt |t|j}tj|j|j d|d Snt d |dS) aA variant of X.getnnz() with extension to weighting on axis 0 Useful in efficiently calculating multilabel metrics. Parameters ---------- X : CSR sparse matrix, shape = (n_samples, n_labels) Input data. axis : None, 0 or 1 The axis on which the data is aggregated. sample_weight : array, shape = (n_samples,), optional Weight for each row of X. rrZcsrz#Expected CSR sparse format, got {0}N) minlength)rLweightszUnsupported axis: {0}) rr rCrdotrrZbincountrrrr)r rZ sample_weightoutrMr r r count_nonzeros,   rPcCspt||}|stjSt|dk}t|d\}}||rLt||||St|d|||t||||dS)zCompute the median of data with n_zeros additional zeros. This function is used to support sparse matrices; it modifies data in-place rr1rg@)r;rnanrPdivmodsort_get_elem_at_rank)rn_zerosZn_elems n_negativeZmiddleZis_oddr r r _get_medians rWcCs,||kr||S|||kr dS|||S)z@Find the value in data augmented with n_zeros for the given rankrr )ZrankrrVrUr r r rTs  rTc Cst|tjstd|j|j}|j\}}t|}xZt t |dd|ddD]8\}\}}t |j ||}||j } t|| ||<qTW|S)a'Find the median across axis 0 of a CSC matrix. It is equivalent to doing np.median(X, axis=0). Parameters ---------- X : CSC sparse matrix, shape (n_samples, n_features) Input data. Returns ------- median : ndarray, shape (n_features,) Median. z%Expected matrix of CSC format, got %sNrJr)rrr#r rrrrr: enumeratezipcopyrsizerW) r rZ n_samplesZ n_featuresZmedianZf_indstartendrZnzr r r csc_median_axis_0s   , r^)F)NN)"Z scipy.sparseZsparserZnumpyrZsparsefuncs_fastrr rr!rr(rrrrr$r)r*r+r0r2r3r4r7rBrFrGrHrIrPrWrTr^r r r r s0 %?7 # 0