c?0dZddlZddlZddlZddlZddlZddlZddlm Z ddl m Z m Z m Z ddlmZddlmZejeZdZdZdd Zdd Zdd ZGdde jZGdde jejZddZdddej ddfdZ!dS)av Module for `Latent Semantic Analysis (aka Latent Semantic Indexing) `_. Implements fast truncated SVD (Singular Value Decomposition). The SVD decomposition can be updated with new observations at any time, for an online, incremental, memory-efficient training. This module actually contains several algorithms for decomposition of large corpora, a combination of which effectively and transparently allows building LSI models for: * corpora much larger than RAM: only constant memory is needed, independent of the corpus size * corpora that are streamed: documents are only accessed sequentially, no random access * corpora that cannot be even temporarily stored: each document can only be seen once and must be processed immediately (one-pass algorithm) * distributed computing for very large corpora, making use of a cluster of machines Wall-clock `performance on the English Wikipedia `_ (2G corpus positions, 3.2M documents, 100K features, 0.5G non-zero entries in the final TF-IDF matrix), requesting the top 400 LSI factors: ====================================================== ============ ================== algorithm serial distributed ====================================================== ============ ================== one-pass merge algorithm 5h14m 1h41m multi-pass stochastic algo (with 2 power iterations) 5h39m N/A [1]_ ====================================================== ============ ================== *serial* = Core 2 Duo MacBook Pro 2.53Ghz, 4GB RAM, libVec *distributed* = cluster of four logical nodes on three physical machines, each with dual core Xeon 2.0GHz, 4GB RAM, ATLAS Examples -------- .. sourcecode:: pycon >>> from gensim.test.utils import common_dictionary, common_corpus >>> from gensim.models import LsiModel >>> >>> model = LsiModel(common_corpus, id2word=common_dictionary) >>> vectorized_corpus = model[common_corpus] # vectorize input copus in BoW format .. [1] The stochastic algo could be distributed too, but most time is already spent reading/decompressing the input from disk in its 4 passes. The extra network traffic due to data distribution across cluster nodes would likely make it *slower*. N) sparsetools) interfacesmatutilsutils) basemodel)is_emptydMbP?c ptjdtj|tj|z z }dt tj|t |d|z kdz}t ||}td|d||dz z|S)aFind how many factors should be kept to avoid storing spurious (tiny, numerically unstable) values. Parameters ---------- s : list of float Eigenvalues of the original matrix. k : int Maximum desired rank (number of factors) discard: float Percentage of the spectrum's energy to be discarded. Returns ------- int Rank (number of factors) of the reduced matrix. ?rz9keeping %i factors (discarding %.3f%% of energy spectrum)r ) npabscumsumsumlenwhereminloggerinfo)skdiscard rel_spectrumsmalls 6lib/python3.11/site-packages/gensim/models/lsimodel.py clip_spectrumrRs(6# !bfQii- 8 8899L BH\Cq,A,AABB1EFF FE Au A KKKQPSVbcdghchViPijjj Hc|jjs5td|j|t j|}|S)a/Get an array laid out in Fortran order in memory. Parameters ---------- a : numpy.ndarray Input array. name : str, optional Array name, used only for logging purposes. Returns ------- np.ndarray The input `a` in Fortran, or column-major order. z'converting %s array %s to FORTRAN order)flags f_contiguousrdebugshaperasfortranarrayanames rasfarrayr*ns? 7 ! >NNN  a  Hrc|jjs5td|j|t j|}|S)a;Return a contiguous array in memory (C order). Parameters ---------- a : numpy.ndarray Input array. name : str, optional Array name, used for logging purposes. Returns ------- np.ndarray Contiguous array (row-major order) of same shape and content as `a`. z!converting %s array %s to C order)r" contiguousrr$r%rascontiguousarrayr's rascarrayr.s? 7 $ 8!'4HHH   # # Hrc>eZdZdZddeeejdfdZdZ ddZ dS) ProjectionaLow dimensional projection of a term-document matrix. This is the class taking care of the 'core math': interfacing with corpora, splitting large corpora into chunks and merging them etc. This done through the higher-level :class:`~gensim.models.lsimodel.LsiModel` class. Notes ----- The projection can be later updated by merging it with another :class:`~gensim.models.lsimodel.Projection` via :meth:`~gensim.models.lsimodel.Projection.merge`. This is how incremental training actually happens. NFc ||c|_|_||_||_||_|?|s4t ||t j||j|j||j\} } n ddl} n#t$rtdwxYwt dt|j tj|st#j|}| ||dz\} } } | j} ~ ~ t)| dz|j}| ddd|f|_| d||_dSd\|_|_dS) a/Construct the (U, S) projection from a corpus. Parameters ---------- m : int Number of features (terms) in the corpus. k : int Desired rank of the decomposed matrix. docs : {iterable of list of (int, float), scipy.sparse.csc} Corpus in BoW format or as sparse matrix. use_svdlibc : bool, optional If True - will use `sparsesvd library `_, otherwise - our own version will be used. power_iters: int, optional Number of power iteration steps to be used. Tune to improve accuracy. extra_dims : int, optional Extra samples to be used besides the rank `k`. Tune to improve accuracy. dtype : numpy.dtype, optional Enforces a type for elements of the decomposed matrix. random_seed: {None, int}, optional Random seed used to initialize the pseudo-random number generator, a local instance of numpy.random.RandomState instance. N) chunksize num_terms power_iters extra_dimsdtype random_seedrzH`sparsesvd` module requested but not found; run `easy_install sparsesvd`z!computing sparse SVD of %s matrixr NN)mrr4r5r7stochastic_svdsysmaxsize sparsesvd ImportErrorrrstrr%scipysparseissparser corpus2cscTrcopyur)selfr:rdocs use_svdlibcr4r5r6r7rGrr>utvts r__init__zProjection.__init__s8A&$&  ( 2%!s{T-=#eIY[[[11 r$$$$$"rrr%&pqqqr ?TZQQQ|,,T225#.t44D%//a"f== ArD!!q&$&11qqq"1"uX]]__DFrrUZZ\\DFFF'NDFDFFFs A$$A>c\t|j|j|j|j|jS)zGet an empty Projection with the same parameters as the current object. Returns ------- :class:`~gensim.models.lsimodel.Projection` An empty copy (without corpus) of the current projection. )r4r5r7)r0r:rr4r5r7rHs r empty_likezProjection.empty_likes3 FDF(8D4D    rr c 6 |jdS|j>|j|_|j|_dS|j|jkrt d|jd|jt dt|jjt|jj|jjd|jjd|jjd}}}t dt|jd|_tj |jj |j}t|jd|_|xjtj |j|zc_|jg|_tj|j\}}|jrJtjtj||jztj||jgtjtjgddt/|||tj||jgg} t d | j t0j| d \} } } n#t0jj$rnt d t0jtj | | j d \} } } tj| } YnwxYwt=| d z|j} tj| d|d| ftj| |dd| f| d| } }} t d| |_t|jd|_tj |j| |_t|d}tj ||}|xj|z c_|jjddkrNtA|jjdD]0}|jd|fdkr|jdd|fxxdzcc</dSdS)aMerge current :class:`~gensim.models.lsimodel.Projection` instance with another. Warnings -------- The content of `other` is destroyed in the process, so pass this function a copy of `other` if you need it further. The `other` :class:`~gensim.models.lsimodel.Projection` is expected to contain the same number of features. Parameters ---------- other : :class:`~gensim.models.lsimodel.Projection` The Projection object to be merged into the current one. It will be destroyed after merging. decay : float, optional Weight of existing observations relatively to new ones. Setting `decay` < 1.0 causes re-orientation towards new data trends in the input document stream, by giving less emphasis to old observations. This allows LSA to gradually "forget" old observations (documents) and give more preference to new ones. Nz'vector space mismatch: update is using z features, expected zmerging projections: %s + %srrz!constructing orthogonal componentzself.uz computing SVD of %s dense matrixF full_matricesz"SVD(A) failed; trying SVD(A * A^T)r zupdating orthonormal basis Uq)!rGrFrr: ValueErrorrrr@r%r$r*rdotrEr.r qr_destroybmatdiagmultiplypadarrayreshaperrAlinalgsvd LinAlgErrorerrorsqrtrrrange)rHotherdecayr:n1n2crTrru_ks_k_u1_ku2_kis rmergezProjection.merges( 7  F 6 W\\^^DFW\\^^DF F 6UW  *UZU\U\U\^b^d^de   2C 4E4Es57=GYGYZZZFLOTV\!_egmA6Fr2  8999$&(++ F468UW % %$&(++ 26$&!$$$7)"57++17 G WUTV^ $ $bk!UW&=&= > \"(2,,..q!44c!Rjj" E Er{STV[V]G^G^ _     7AAA  ,**1E*BBKCaa|'    LL= > > >,**26!QS>>*OOKCa'#,,CCC   #(DF + +(3ssBQBw<00"(3rssBQBw<2H2H#bqb'Cd  3444$&(++%% Q   F1dOO !  6<?Q  )46<?++ ) )6!Q$<#%)F111a4LLLD(LLL ) ) ) )s%J==A?L?>L?r ) __name__ __module__ __qualname____doc__P2_EXTRA_ITERS P2_EXTRA_DIMSrfloat64rMrPrrrrr0r0sn  "u.$BJD8(8(8(8(t    Z)Z)Z)Z)Z)Z)rr0c eZdZdZdddddddeeejdf dZdd Z d Z dd Z d Z ddZ ddZddZfdZefdZxZS)LsiModela[Model for `Latent Semantic Indexing `_. The decomposition algorithm is described in `"Fast and Faster: A Comparison of Two Streamed Matrix Decomposition Algorithms" `_. Notes ----- * :attr:`gensim.models.lsimodel.LsiModel.projection.u` - left singular vectors, * :attr:`gensim.models.lsimodel.LsiModel.projection.s` - singular values, * ``model[training_corpus]`` - right singular vectors (can be reconstructed if needed). See Also -------- `FAQ about LSI matrices `_. Examples -------- .. sourcecode:: pycon >>> from gensim.test.utils import common_corpus, common_dictionary, get_tmpfile >>> from gensim.models import LsiModel >>> >>> model = LsiModel(common_corpus[:3], id2word=common_dictionary) # train model >>> vector = model[common_corpus[4]] # apply model to BoW document >>> model.add_documents(common_corpus[4:]) # update model with new documents >>> tmp_fname = get_tmpfile("lsi.model") >>> model.save(tmp_fname) # save model >>> loaded_model = LsiModel.load(tmp_fname) # load model N Nr FTc ||_t||_t||_t ||_|r|st dd}||_| |c|_ |_ | |_ | |_ ||jtd|jMt dtj||_t!|j|_n7d|jr&t%|jndz|_d|_t+|j|j|j |j | |j |_d|_|s#t d d|_n|st5d  ddl} | d } t d t=| j| |j||||j |j d|| |_t!| !|_t d|jn?#tD$r2}t #d|tId|zd}~wwxYw|]tKj%}|&||'dd|dtKj%|z dddSdS)aBuild an LSI model. Parameters ---------- corpus : {iterable of list of (int, float), scipy.sparse.csc}, optional Stream of document vectors or a sparse matrix of shape (`num_documents`, `num_terms`). num_topics : int, optional Number of requested factors (latent dimensions) id2word : dict of {int: str}, optional ID to word mapping, optional. chunksize : int, optional Number of documents to be used in each training chunk. decay : float, optional Weight of existing observations relatively to new ones. distributed : bool, optional If True - distributed mode (parallel execution on several machines) will be used. onepass : bool, optional Whether the one-pass algorithm should be used for training. Pass `False` to force a multi-pass stochastic algorithm. power_iters: int, optional Number of power iteration steps to be used. Increasing the number of power iterations improves accuracy, but lowers performance extra_samples : int, optional Extra samples to be used besides the rank `k`. Can improve accuracy. dtype : type, optional Enforces a type for elements of the decomposed matrix. random_seed: {None, int}, optional Random seed used to initialize the pseudo-random number generator, a local instance of numpy.random.RandomState instance. z2forcing the one-pass algorithm for distributed LSATNzYat least one of corpus/id2word must be specified, to establish input space dimensionalityzHno word id mapping provided; initializing from corpus, assuming identityrr)r4r5r6r7z%using serial LSI version on this nodezfdistributed stochastic LSA not implemented yet; run either distributed one-pass, or serial randomized.zPYRONAME:gensim.lsi_dispatcherzlooking for dispatcher at %sF)id2word num_topicsr2rgr4 extra_samples distributedonepassz)using distributed version with %i workersz)failed to initialize distributed LSI (%s)createdztrained z in z.2fr)msg)(rintrr2floatrgrwarningrrr4r6r7rWrdict_from_corpusrr3maxkeysdocs_processedr0 projection numworkersr dispatcherNotImplementedErrorPyro4Proxyr$r@_pyroUri initialize getworkers Exceptionrc RuntimeErrortime add_documentsadd_lifecycle_event)rHcorpusrrr2rgrrr4rr6r7rrerrstarts rrMzLsiModel.__init__ns+H j//Y5\\   STTT /"> GYYYY V V V H#NNN"#NQT#TUUU V  IKKE   v & & &  $ $CtCCu)<CCCC %       s#B?I## J--JJc td||j}||j}t |rtdt j|s|j st|j |j d|j |j}t||j |j ||j|j|j |j\|_|_|j|||xjt/|drt1|ndz c_dSd}|jr9td |j|jt9t;j||D]o\}}td t?d |D}t d tCj"|t1||j ||j } ~|| j#dz }|jrRt d||j$| ~ td|t|j |j | |j|j|j |j}~ |j||~td||%dq|jr8td|j&|_|xj|z c_dS|jr Jdt|j |j |'|j|j|j }|j||td|j#d|xj|j#dz c_dS)a_Update model with new `corpus`. Parameters ---------- corpus : {iterable of list of (int, float), scipy.sparse.csc} Stream of document vectors or sparse matrix of shape (`num_terms`, num_documents). chunksize : int, optional Number of documents to be used in each training chunk, will use `self.chunksize` if not specified. decay : float, optional Weight of existing observations relatively to new ones, will use `self.decay` if not specified. Notes ----- Training proceeds in chunks of `chunksize` documents at a time. The size of `chunksize` is a tradeoff between increased speed (bigger `chunksize`) vs. lower memory footprint (smaller `chunksize`). If the distributed mode is on, each chunk is sent to a different worker/computer. z!updating model with new documentsNzLLsiModel.add_documents() called but no documents provided, is this intended?)r6r7)r3r2r5r4r6r7)rg__len__rzinitializing %s workersz"preparing a new chunk of documentsc34K|]}t|VdSNr.0docs r z)LsiModel.add_documents..s(883c#hh888888rzconverting corpus to csc format)num_docsr3num_nnzr6rzcreating job #%izdispatched documents up to #%s)r5r4r6r7zprocessed documents up to #%szFreached the end of input; now waiting for all remaining jobs to finishz&must be in serial mode to receive jobs)r5r4r6z$processed sparse job of %i documents)(rrr2rgrrrArBrCrr0r3rr6r7r;rr4rGrrrrrhasattrrrrreset enumeratergrouperrr$rrDr%putjob print_topicsgetstatetocsc) rHrr2rgupdatedoc_nochunk_nochunknnzjobs rrzLsiModel.add_documentss&  7888  'I  JE F   k NNi j j j|$$V,,A 3<7 .#NDOT*$2B&4DO"n #1t?OW[Wa $ 0 &&&"&( %%fE%:::##gfi6P6P'Ws6{{{VWW####?,KK 94?KKKO))+++'0vy1Q1Q'R'R--OHeKK DEEE88%88888CLL!BCCC"-E dn #4:Ccil*F- %7BBB..s333 $DfMMMM", NDOSTM_(,(8 X\Xh""" --fE-BBB" $CVLLL))!,,,,?AKK hiii&*o&>&>&@&@DO##v-#### P P(P P P PDL^ ,DJF O ! !& ! 6 6 6 KK> Q P P P   6<? 2    rc `|jjd|jd|jd|jd|jd S)zGet a human readable representation of model. Returns ------- str A human readable string of the current objects parameters. z ) __class__rtr3rrgr2rOs r__str__zLsiModel.__str__2s@ N # # #T^^^T___djjjRVR`R`R`  rc|jjtdtj|\}}|r|r|||S|s|g}t j||j|jjj }|j |jjddd|j fzj }|s| d}|rd|jj d|j z |z}|st j|}nt j|}|S)aGet the latent representation for `bow`. Parameters ---------- bow : {list of (int, int), iterable of list of (int, int)} Document or corpus in BoW representation. scaled : bool, optional If True - topics will be scaled by the inverse of singular values. chunksize : int, optional Number of documents to be used in each applying chunk. Returns ------- list of (int, float) Latent representation of topics in BoW format for document **OR** :class:`gensim.matutils.Dense2Corpus` Latent representation of corpus in BoW format if `bow` is corpus. Nz9No training data provided - LSI model not initialized yet)r2r3r6rr )rrGrWr is_corpus_applyrrDr3r6rErr_r full2sparse Dense2Corpus)rHbowscaledr2rvec topic_distresults r __getitem__zLsiModel.__getitem__?s3( ?  ZXYY Y-- 3  9 9 ;;si;88 8 %C!#tGXG^___edo/3CDO3C0CDDG  0#++B//J  R 12B4?2B CCzQJ 7)*55FF*:66F rc |jjj}t|}g}t |D]}t j||ddf}t jt j t j ||}| d|z|z t j |S)aGet the topic vectors. Notes ----- The number of topics can actually be smaller than `self.num_topics`, if there were not enough factors in the matrix (real rank of input matrix smaller than `self.num_topics`). Returns ------- np.ndarray The term topic matrix with shape (`num_topics`, `vocabulary_size`) Nr ) rrGrErrerasarrayflattenrdrrXappendr^)rH projectionsrtopicsrqrjnorms r get_topicszLsiModel.get_topicsso') %% z"" * *A ;q!!!t,--5577A726"&A,,//00D MM#'D. ) ) ) )xr c|tjjjkrdSt jjjj|ddft jt jt j tj t j |d}fd|DS)aQGet the words that define a topic along with their contribution. This is actually the left singular vector of the specified topic. The most important words in defining the topic (greatest absolute value) are included in the output, along with their contribution to the topic. Parameters ---------- topicno : int The topics id number. topn : int Number of words to be included to the result. Returns ------- list of (str, float) Topic representation in BoW format. r NT)reversecXg|]&}|jv j|d|zz f'Srs)r)rvalrjrrHs r z'LsiModel.show_topic..s@```SCSWS_L_`c"C!C&L4$78```r) rrrGrErrrrdrrXrargsortr)rHtopicnotopnmostrjrs` @@r show_topiczLsiModel.show_topics0 c$/+-.. . 2 Jt(*7AAA:6 7 7 ? ? A AwrvbfQll++,,q 4>>>a`````D````rrcg}|dkr|j}tt||jD]}|t|jjkrw|r|||}n|||}|||f|r-t d||jj|||S)a=Get the most significant topics. Parameters ---------- num_topics : int, optional The number of topics to be selected, if -1 - all topics will be in result (ordered by significance). num_words : int, optional The number of words to be included per topics (ordered by significance). log : bool, optional If True - log topics with logger. formatted : bool, optional If True - each topic represented as string, otherwise - in BoW format. Returns ------- list of (int, str) If `formatted=True`, return sequence with (topic_id, string representation of topics) **OR** list of (int, list of (str, float)) Otherwise, return sequence with (topic_id, [(word, value), ... ]). r)rztopic #%i(%.3f): %s) rrerrrr print_topicrrrr)rHr num_wordslog formattedshownrqtopics r show_topicszLsiModel.show_topicss, > )Js:t7788 W WA3t())) W? ,,QY,??EE OOAIO>>E aZ(((WKK 5q$/:KA:NPUVVV rrc t|j|jj|jjt t |t|jjj|dS)aPrint (to log) the most salient words of the first `num_topics` topics. Unlike :meth:`~gensim.models.lsimodel.LsiModel.print_topics`, this looks for words that are significant for a particular topic *and* not for others. This *should* result in a more human-interpretable description of topics. Alias for :func:`~gensim.models.lsimodel.print_debug`. Parameters ---------- num_topics : int, optional The number of topics to be selected (ordered by significance). num_words : int, optional The number of words to be included per topics (ordered by significance). )rN) print_debugrrrGrrerrrE)rHrrs rrzLsiModel.print_debugs`$  L$/+T_-> #j#do&7&9":":;; < <      rc|j)|jjtj|dg|Ri|t t |j|g|Rdddgi|dS)aSave the model to a file. Notes ----- Large internal arrays may be stored into separate files, with `fname` as prefix. Warnings -------- Do not save as a compressed file if you intend to load the file back with `mmap`. Parameters ---------- fname : str Path to output file. *args Variable length argument list, see :meth:`gensim.utils.SaveLoad.save`. **kwargs Arbitrary keyword arguments, see :meth:`gensim.utils.SaveLoad.save`. See Also -------- :meth:`~gensim.models.lsimodel.LsiModel.load` N .projectionignorerr)rsaversmart_extensionsuperr})rHfnameargskwargsrs rrz LsiModel.saves4 ? _ DO !6um!L!L ^t ^ ^ ^W] ^ ^ ^"h"5_4___|8T_X^_____rc\|dd|d<tt|j|g|Ri|}t j|d} tt|j|g|Ri||_n-#t$r }tj d||Yd}~nd}~wwxYw|S)agLoad a previously saved object using :meth:`~gensim.models.lsimodel.LsiModel.save` from file. Notes ----- Large arrays can be memmap'ed back as read-only (shared memory) by setting the `mmap='r'` parameter. Parameters ---------- fname : str Path to file that contains LsiModel. *args Variable length argument list, see :meth:`gensim.utils.SaveLoad.load`. **kwargs Arbitrary keyword arguments, see :meth:`gensim.utils.SaveLoad.load`. See Also -------- :meth:`~gensim.models.lsimodel.LsiModel.save` Returns ------- :class:`~gensim.models.lsimodel.LsiModel` Loaded instance. Raises ------ IOError When methods are called on instance (should be called from class). mmapNrz%failed to load projection from %s: %s) getrr}loadrrrrloggingr)clsrrrrprojection_fnameers rrz LsiModel.loads@ FD11v*x%%*5B4BBB6BB 0 FF Z 9h 4 4 9:J \T \ \ \U[ \ \F   Z Z Z OCEUWX Y Y Y Y Y Y Y Y Z s)A?? B) B$$B)r9)Fr)r)rrFT)rr)rtrurvrwrxryrrzrMrrrrrrrr classmethodr __classcell__)rs@rr}r}Ls5D#tu5$N'rztccccJ]3]3]3]3~    DDDDL   .aaaaB""""H    0`````<&&&&[&&&&&rr}rc ||dz }tdt|t|i}}t |D]\}}t jt j|}|t j t j t j ||z } |D]3} | | g | | |f4td||t|D]3} t|| d} | d\} } || | fdkrd }nd }gg}}| D]S\}}|||| fzd kr=| d ||||| ffzt||krnT| D]S\}}|||| fzd kr=| d ||||| ffzt||krnTtd| || d|d|5dS)a[Log the most salient words per topic. Parameters ---------- id2token : :class:`~gensim.corpora.dictionary.Dictionary` Mapping from ID to word in the Dictionary. u : np.ndarray The 2D U decomposition matrix. s : np.ndarray The 1D reduced array of eigenvalues used for decomposition. topics : list of int Sequence of topic IDs to be printed num_words : int, optional Number of words to be included for each topic. num_neg : int, optional Number of words with a negative contribution to a topic that should be included. Nr z+computing word-topic salience for %i topicszprinting %i+%i salient wordsc.t|d S)Nr)r)xs rzprint_debug..css1Q4yyjr)keyrrUrVr g-C6?z%s(%.3f)g-C6ztopic #%s(%.3f): %s, ..., %sz, )rrrsetrrrrrrdrrX setdefaultrr$sortedrjoin)id2tokenrGrrrnum_negruvecnouvecudiffrweightsrnr normalizeposnegweights rrrAs& a- KK=s6{{KKK[["FF!! HH vbj&&..0011rwrvbfT4&8&899::: H HE   eR ( ( / /uv0F G G G G H LL/GDDD &&ee,@,@AAA!*4 T5[>C  IIIrS%  NFF1VU]++f4  :&)91VU];K(LLMMMs88y(E%  NFF1VU]++g5  :&)91VU];K(LLMMMs88w&E 2E1U8TYYs^^UYU^U^_bUcUcdddd-eerrgư>c t|}|tdd|z} n|t|z} td| |z |t|}t j||| f} tdt | jtj |} tj |r^|j\} } || ksJd| d|d | d d | | f| j}t!j| | | |j|j|j|| ~| j|kr| |} td t | j| g} t-j| \}}td |t3|D])}|j|z}||zg}t-j|\}}*n"d}t7t9j||D] \}}td||zt=d|D}t-j|||}|j\} } | |ksJ| |ksJ|| z }td| d d | | f|}t!j| | | |j|j|j|| ~~ | g} t-j| \}}t3|D]}td|dz| }d |dd<t7t9j||D]P\}}td||z|t-j|||}|j|z}||z}~||z }Q~|g}t-j|\}}|ddd| fj }~tj |r[||z}tdt |jtj!"|d\}}}~~nFt j|jd|jdf|}tdt |jt7t9j||D]d\}}td||z|t-j|||j}||z}~|t j#||jz }~etdt |jtj!"|\}}}t j$|}|j }~tdtK|dz||}|ddd|f }|d|}t j#||}||||fS)aRun truncated Singular Value Decomposition (SVD) on a sparse input. Parameters ---------- corpus : {iterable of list of (int, float), scipy.sparse} Input corpus as a stream (does not have to fit in RAM) or a sparse matrix of shape (`num_terms`, num_documents). rank : int Desired number of factors to be retained after decomposition. num_terms : int The number of features (terms) in `corpus`. chunksize : int, optional Number of documents to be used in each training chunk. extra_dims : int, optional Extra samples to be used besides the rank `k`. Can improve accuracy. power_iters: int, optional Number of power iteration steps to be used. Increasing the number of power iterations improves accuracy, but lowers performance. dtype : numpy.dtype, optional Enforces a type for elements of the decomposed matrix. eps: float, optional Percentage of the spectrum's energy to be discarded. random_seed: {None, int}, optional Random seed used to initialize the pseudo-random number generator, a local instance of numpy.random.RandomState instance. Notes ----- The corpus may be larger than RAM (iterator of vectors), if `corpus` is a `scipy.sparse.csc` instead, it is assumed the whole corpus fits into core memory and a different (more efficient) code path is chosen. This may return less than the requested number of top `rank` factors, in case the input itself is of lower rank. The `extra_dims` (oversampling) and especially `power_iters` (power iterations) parameters affect accuracy of the decomposition. This algorithm uses `2 + power_iters` passes over the input data. In case you can only afford a single pass, set `onepass=True` in :class:`~gensim.models.lsimodel.LsiModel` and avoid using this function directly. The decomposition algorithm is based on `"Finding structure with randomness: Probabilistic algorithms for constructing approximate matrix decompositions" `_. Returns ------- (np.ndarray 2D, np.ndarray 1D) The left singular vectors and the singular values of the `corpus`. Nrr z.using %i extra samples and %i power iterations)r6r%z(1st phase: constructing %s action matrixz mismatch in number of features: z in sparse matrix vs. z parameterrUr z!orthonormalizing %s action matrixzrunning %i power iterationsrzPROGRESS: at document #%ic34K|]}t|VdSrrrs rrz!stochastic_svd..s(..CHH......rrzmultiplying chunk * gausszrunning power iteration #%irzPROGRESS: at document #%i/%iz)2nd phase: running dense svd on %s matrixFrR)r%r6z,2nd phase: constructing %s covariance matrixz3running dense decomposition on %s covariance matrixz!computing the final decomposition)r)&rrrrrzerosr@r%random RandomStaterArBrCnormalastyper6r csc_matvecsindptrindicesdataravelrrYr$rerErrrrrDrFr`rarXrdr)rrankr3r2r5r4r6epsr7samplesy random_stater:norTrnrrrr power_iteryoldtmpqtbrGrLrkeeps rr;r;{sh t99D)b!d(##Z( KK@'D.R]^^^II uY$8999A KK:CLLIII9((55L |V$$<*|1A~ppp!pp[dpppppp   S1g, 7 7 > >qw G G q'6=&. KAGGII     7e  A 7QWFFF C"1%%1 2K@@@{## * *A1 A! A&q))DAqq * (vy)I)IJJ  OHe KK3h6J L L L.......A'%PPPE;DAq > ! ! ! > ! ! ! MH LL4 5 5 5##Cq'l==DDUKKA  #1gu|U] AGGIIqwwyy   qq C"1%%1 ,, * *J KK5zA~ F F F6688DAaaaD#,U]69-M-M#N#N  % :Hyr2s44l $$$$$$..........######!!!!!!  8 $ $     8    ,    ,o)o)o)o)o)o)o)o)drrrrrz+Y-Errrj7e7e7e7ev,1TRZTtf,f,f,f,f,f,r