:i8SSKrSSKrSSKrSSKJr SSKJr SSK J r J r SSKJ r SSK JrJr SS KJr \R$"S 5r"S S \5r\S 5r\S5r"SS\5r"SS\5r"SS\5rg)N)njit)utils) Deserializer Serializer) MaskedModel)DimensionErrorInvalidClusteringError)Maskershapc\^\rSrSrSrS SjrSrSrU4Sjr\ S U4Sjj5r Sr U=r $) Tabularz2A common base class for Independent and Partition.cSUl[U[R5(a$URUlUR nSUl[U[5(aNSU;aHURSS5Ul URSS5Ul [R"USS5n[US5(a)URSU:a[R "X5nXlX0lX@lX lUbUb [+S5eUbh[U[,5(a[R."XS 9UlO1[U[R05(aX0lO [3S 5eSUlOUbX@lSUlOSUlSUlUR55Ul[R8"URS [:S 9UlUR"RUlSUlg) aNThis masks out tabular features by integrating over the given background dataset. Parameters ---------- data : np.array, pandas.DataFrame The background dataset that is used for masking. max_samples : int The maximum number of samples to use from the passed background data. If data has more than max_samples then shap.utils.sample is used to subsample the dataset. The number of samples coming out of the masker (to be integrated over) matches the number of samples in the background dataset. This means larger background dataset cause longer runtimes. Normally about 1, 10, 100, or 1000 background samples are reasonable choices. clustering : string or None (default) or numpy.ndarray The distance metric to use for creating the clustering of the features. The distance function can be any valid scipy.spatial.distance.pdist's metric argument. However we suggest using 'correlation' in most cases. The full list of options is `braycurtis`, `canberra`, `chebyshev`, `cityblock`, `correlation`, `cosine`, `dice`, `euclidean`, `hamming`, `jaccard`, `jensenshannon`, `kulsinski`, `mahalanobis`, `matching`, `minkowski`, `rogerstanimoto`, `russellrao`, `seuclidean`, `sokalmichener`, `sokalsneath`, `sqeuclidean`, `yule`. These are all the options from scipy.spatial.distance.pdist's metric argument. FTmeanNcovrshapezKYou cannot pass both 'clustering' and 'partition'. Please provide only one.)metriczoUnknown clustering given! Make sure you pass a distance metric as a string, or a clustering as a numpy.ndarray.r dtype) output_dataframe isinstancepd DataFramecolumns feature_namesvaluesdictgetrrnp expand_dimshasattrrrsampledata clustering partition max_samples ValueErrorstrhclustndarrayr copy _masked_datazerosbool _last_masksupports_delta_masking)selfr%r(r&r's X/srv/projetos/modelo_ml_acdoc/venv/lib/python3.13/site-packages/shap/maskers/_tabular.py__init__Tabular.__init__s4!& dBLL ) )!%D ;;D$(D ! dD ! !fn.DIxxt,DH>>$v,2D 4 ! !djjmk&A<<2D $"&  !i&;jk k  #*c**"',,t"GJ 33",,F"DN  "&N"DO"DO!DN!IIK((4::a==YY__ &*#c URX5n[UR5S:wd*URSURRS:wa [ S5e[ R "UR[ R5(GaAURU5)n[ R"[U5[S9nUS:R5n[ R"XPRS4[S9n[ R"XPRS-URS45nSURSS&URURSS&[!UUUUURURURUU["R$5 UR&(a![(R*"XpR,S94U4$U4U4$X!-UR[ R."U5--URSS&XRSS&UR&(a)[(R*"URUR,S9$UR4$)Nr rzNThe input passed for tabular masking does not match the background data shape!rFr)_standardize_masklenrr%r r! issubdtyperinteger invariantsr/intsumr0r1r._delta_maskingrdelta_mask_noop_valuerrrrinvert)r3maskxvariantscurr_delta_inds num_masksvarying_rows_outmasked_inputs_outs r4__call__Tabular.__call__ds%%d. qww<1  diiooa.@ @ !qr r ==RZZ 0 0**H hhs4yTabular.invariantss 77diiooab) ) pagg,HIdiiooab)*+  zz!YY''r7c R>[TU]U5 [USSS9nUR(a:URS[R "UR URS95 OS[USS5b)URSURUR45 OURSUR 5 URSUR5 URS UR5 URS UR5 SSS5 g!,(df  g=f) z(Write a Tabular masker to a file stream.shap.maskers.Tabularr)versionr%r9rNr(r&r')supersaverrrrr%rgetattrrrr(r&r')r3out_files __class__s r4rU Tabular.saves  X"8! D$$vr||DIIt?Q?QRSvt,8v 48845vtyy) FF=$"2"2 3 FF< 1 FF; /E D Ds C3D D&cB>U(aURU5$[TU] USS9n[USSSS9nURS5US'URS5US'URS5US'URS 5US 'S S S 5 U$!,(df  U$=f) z)Load a Tabular masker from a file stream.F) instantiaterRr) min_version max_versionr%r(r&r'N)_instantiated_loadrTloadr)clsin_filer\kwargsrXrYs r4r` Tabular.loads ))'2 2g59 '#9qVW X\]VVF^F6N$%FF=$9F= !#$66,#7F< "#&&"5F;  Y  Y X  s AB B) r1r.r&rr%rr(rrr'rr2)dNN)T) __name__ __module__ __qualname____firstlineno____doc__r5rKr>rU classmethodr`__static_attributes__ __classcell__rYs@r4rrs1<J+b($d($0$  r7rcjX:XagX (aUSS2U4USS2U4'SX 'gX@USS2U4'SX 'g)NFT)dind masked_inputs last_maskr%rE noop_codes r4_single_delta_maskrusD  !%ag ag !" ag r7c Sn Sn Sn Sn URSnU [U5:aU S- n Sn X US'X*S:a2X**S- X*'[X*XEXaU 5 U S- n X U -X*'X*S:aM2[X*XEXaU 5 XHXU-&XS-- n U S:Xa SX;SS24'OCU S:XaUSS2X*4X;SS24'O+[R"USS2USU S-4SS9S:X;SS24'X- n U [U5:aMgg)zImplements the special (high speed) delta masking API that only flips the positions we need to. Note that we attempt to avoid doing any allocation inside this function for speed reasons. rr TN)axis)rr;rur!r@)masksrErGrImasked_inputs_tmprsr%rFrJrtdposi masks_pos output_posNs r4rArAs_" D AIJ"A c%j  Q"-#a'%4%:$:Q$>O ! 46GTX]f g AID$)d*:$;O ! #a' ?02CPTYbc:K*A~6AX  6%) T "qy)1!_5J2J)K A&*,OJdUVh[TU]XSS9 g)aBuild a Independent masker with the given background data. Parameters ---------- data : numpy.ndarray, pandas.DataFrame The background dataset that is used for masking. max_samples : int The maximum number of samples to use from the passed background data. If data has more than max_samples then shap.utils.sample is used to subsample the dataset. The number of samples coming out of the masker (to be integrated over) matches the number of samples in the background dataset. This means larger background dataset cause longer runtimes. Normally about 1, 10, 100, or 1000 background samples are reasonable choices. N)r(r&rTr5)r3r%r(rYs r4r5Independent.__init__s 4Hr7rp)rerfrgrhrirjr5rlrmrns@r4rr s[IIr7rc0^\rSrSrSrSU4SjjrSrU=r$) Partitioni"zThis masks out tabular features by integrating over the given background dataset. Unlike Independent, Partition respects a hierarchical structure of the data. c$>[TU]XUSS9 g)aBuild a Partition masker with the given background data and clustering. Parameters ---------- data : numpy.ndarray, pandas.DataFrame The background dataset that is used for masking. max_samples : int The maximum number of samples to use from the passed background data. If data has more than max_samples then shap.utils.sample is used to subsample the dataset. The number of samples coming out of the masker (to be integrated over) matches the number of samples in the background dataset. This means larger background dataset cause longer runtimes. Normally about 1, 10, 100, or 1000 background samples are reasonable choices. clustering : string or numpy.ndarray If a string, then this is the distance metric to use for creating the clustering of the features. The distance function can be any valid scipy.spatial.distance.pdist's metric argument. However we suggest using 'correlation' in most cases. The full list of options is `braycurtis`, `canberra`, `chebyshev`, `cityblock`, `correlation`, `cosine`, `dice`, `euclidean`, `hamming`, `jaccard`, `jensenshannon`, `kulsinski`, `mahalanobis`, `matching`, `minkowski`, `rogerstanimoto`, `russellrao`, `seuclidean`, `sokalmichener`, `sokalsneath`, `sqeuclidean`, `yule`. These are all the options from scipy.spatial.distance.pdist's metric argument. If an array, then this is assumed to be the clustering of the features. N)r(r&r'r)r3r%r(r&rYs r4r5Partition.__init__(s6 :Y]^r7rp)re correlationrrns@r4rr"s __r7rc"\rSrSrSrSSjrSrg)ImputeiFzThis imputes the values of missing features using the values of the observed features. Unlike Independent, Gaussian imputes missing values based on correlations with observed data points. cU[LaNSU;aHURSS5UlURSS5Ul[R "USS5nXlX lg)zBuild a Partition masker with the given background data and clustering. Parameters ---------- data : numpy.ndarray, pandas.DataFrame or {"mean: numpy.ndarray, "cov": numpy.ndarray} dictionary The background dataset that is used for masking. rNrr)rr rrr!r"r%method)r3r%rs r4r5Impute.__init__LsS 4>$v,2D  r7)rr%rrN)linear)rfrgrhrirjr5rlrpr7r4rrFs  r7r)loggingnumpyr!pandasrnumbarr _serializablerrrutils._exceptionsr r _maskerr getLoggerlogrrurArrrrpr7r4rs4F vfvr33lI'I,!_!_HVr7