Search Results

Search found 15 results on 1 pages for 'svd'.

Page 1/1 | 1 

  • OpenCV SVD Matrix format.

    - by Rick
    I currently have a set of 2D Cartesian coordinates e.g. {(1,3), (2,2), (3,4)} Which will be put into a 2D array, to perform SVD properly would the matrix be put together such that the coordinates form the columns or the rows e.g. 1 3 2 2 3 4 or 1 2 3 3 2 4 I have been doing a little trial and error comparing to examples of SVD I have found online, the resulting matrix usually seems to be negated, with some of the values shuffled around. To clarify further if I had a matrix E which was MxN as shown here http://upload.wikimedia.org/wikipedia/commons/b/bb/Matrix.svg To define the matrix as a 2D array would it be Array[M][N] or Array[N][M] I am assuming this actually matters due to matrix arithmetic not being commutative? Can anyone actually verify this?

    Read the article

  • Performing a SVD on tweets. Memory problem

    - by plotti
    I have generated a huge csv file as an output from my pos tagging and stemming. It looks like this: word1, word2, word3, ..., word14400 person1 1 2 0 1 person2 0 0 1 0 ... person650 It contains the word counts for each person. Like this I am getting characteristic vectors for each person. I want to run a SVD on this beast, but it seems the matrix is too big to be held in memory to perform the operation. My quesion is: should i reduce the column size by removing words which have a column sum of for example 1, which means that they have been used only once. Do I bias the data too much with this attempt? I tried the rapidminer attempt, by loading the csv into the db. and then sequentially reading it in with batches for processing, like rapidminer proposes. But Mysql can't store that many columns in a table. If i transpose the data, and then retranspose it on import it also takes ages.... -- So in general I am asking for advice how to perform a svd on such a corpus.

    Read the article

  • SVD factorization

    - by Simon
    How can i use the bidiagonal and diagonal form to compute the SVD factorization. I think my real problem is that i don't understand this concepts.

    Read the article

  • Matlab Image watermarking question , using both SVD and DWT

    - by Georgek
    Hello all . here is a code that i got over the net ,and it is supposed to embed a watermark of size(50*20) called _copyright.bmp in the Code below . the size of the cover object is (512*512), it is called _lena_std_bw.bmp.What we did here is we did DWT2 2 times for the image , when we reached our second dwt2 cA2 size is 128*128. You should notice that the blocksize and it equals 4, it is used to determine the max msg size based on cA2 according to the following code:max_message=RcA2*CcA2/(blocksize^2). in our current case max_message would equal 128*128/(4^2)=1024. i want to embed a bigger watermark in the 2nd dwt2 and lets say the size of that watermark is 400*10(i can change the dimension using MS PAINT), what i have to do is change the size of the blocksize to 2. so max_message=4096.Matlab gives me 3 errors and they are : ??? Error using == plus Matrix dimensions must agree. Error in == idwt2 at 93 x = upsconv2(a,{Lo_R,Lo_R},sx,dwtEXTM,shift)+ ... % Approximation. Error in == two_dwt_svd_low_low at 88 CAA1 = idwt2(cA22,cH2,cV2,cD2,'haar',[RcA1,CcA1]); The origional Code is (the origional code where blocksize =4): %This algorithm makes DWT for the whole image and after that make DWT for %cH1 and make SVD for cH2 and embed the watermark in every level after SVD %(1) -------------- Embed Watermark ------------------------------------ %Add the watermar W to original image I and give the watermarked image in J %-------------------------------------------------------------------------- % set the gain factor for embeding and threshold for evaluation clc; clear all; close all; % save start time start_time=cputime; % set the value of threshold and alpha thresh=.5; alpha =0.01; % read in the cover object file_name='_lena_std_bw.bmp'; cover_object=double(imread(file_name)); % determine size of watermarked image Mc=size(cover_object,1); %Height Nc=size(cover_object,2); %Width % read in the message image and reshape it into a vector file_name='_copyright.bmp'; message=double(imread(file_name)); T=message; Mm=size(message,1); %Height Nm=size(message,2); %Width % perform 1-level DWT for the whole cover image [cA1,cH1,cV1,cD1] = dwt2(cover_object,'haar'); % determine the size of cA1 [RcA1 CcA1]=size(cA1) % perform 2-level DWT for cA1 [cA2,cH2,cV2,cD2] = dwt2(cA1,'haar'); % determine the size of cA2 [RcA2 CcA2]=size(cA2) % set the value of blocksize blocksize=4 % reshape the watermark to a vector message_vector=round(reshape(message,Mm*Nm,1)./256); W=message_vector; % determine maximum message size based on cA2, and blocksize max_message=RcA2*CcA2/(blocksize^2) % check that the message isn't too large for cover if (length(message) max_message) error('Message too large to fit in Cover Object') end %----------------------- process the image in blocks ---------------------- x=1; y=1; for (kk = 1:length(message_vector)) [cA2u cA2s cA2v]=svd(cA2(y:y+blocksize-1,x:x+blocksize-1)); % if message bit contains zero, modify S of the original image if (message_vector(kk) == 0) cA2s = cA2s*(1 + alpha); % otherwise mask is filled with zeros else cA2s=cA2s; end cA22(y:y+blocksize-1,x:x+blocksize-1)=cA2u*cA2s*cA2v; % move to next block of mask along x; If at end of row, move to next row if (x+blocksize) >= CcA2 x=1; y=y+blocksize; else x=x+blocksize; end end % perform IDWT CAA1 = idwt2(cA22,cH2,cV2,cD2,'haar',[RcA1,CcA1]); watermarked_image= idwt2(CAA1,cH1,cV1,cD1,'haar',[Mc,Nc]); % convert back to uint8 watermarked_image_uint8=uint8(watermarked_image); % write watermarked Image to file imwrite(watermarked_image_uint8,'dwt_watermarked.bmp','bmp'); % display watermarked image figure(1) imshow(watermarked_image_uint8,[]) title('Watermarked Image') %(2) ---------------------------------------------------------------------- %---------- Extract Watermark from attacked watermarked image ------------- %-------------------------------------------------------------------------- % read in the watermarked object file_name='dwt_watermarked.bmp'; watermarked_image=double(imread(file_name)); % determine size of watermarked image Mw=size(watermarked_image,1); %Height Nw=size(watermarked_image,2); %Width % perform 1-level DWT for the whole watermarked image [ca1,ch1,cv1,cd1] = dwt2(watermarked_image,'haar'); % determine the size of ca1 [Rca1 Cca1]=size(ca1); % perform 2-level DWT for ca1 [ca2,ch2,cv2,cd2] = dwt2(ca1,'haar'); % determine the size of ca2 [Rca2 Cca2]=size(ca2); % process the image in blocks % for each block get a bit for message x=1; y=1; for (kk = 1:length(message_vector)) % sets correlation to 1 when patterns are identical to avoid /0 errors % otherwise calcluate difference between the cover image and the % watermarked image [cA2u cA2s cA2v]=svd(cA2(y:y+blocksize-1,x:x+blocksize-1)); [ca2u1 ca2s1 ca2v1]=svd(ca2(y:y+blocksize-1,x:x+blocksize-1)); correlation(kk)=diag(ca2s1-cA2s)'*diag(ca2s1-cA2s)/(alpha*alpha)/(diag(cA2s)*diag(cA2s)); % move on to next block. At and of row move to next row if (x+blocksize) >= Cca2 x=1; y=y+blocksize; else x=x+blocksize; end end % if correlation exceeds average correlation correlation(kk)=correlation(kk)+mean(correlation(1:Mm*Nm)); for kk = 1:length(correlation) if (correlation(kk) > thresh*alpha);%thresh*mean(correlation(1:Mo*No))) message_vector(kk)=0; end end % reshape the message vector and display recovered watermark. figure(2) message=reshape(message_vector(1:Mm*Nm),Mm,Nm); imshow(message,[]) title('Recovered Watermark') % display processing time elapsed_time=cputime-start_time, please do help,its my graduation project and i have been trying this code for along time but failed miserable. Thanks in advance

    Read the article

  • java jama array problem

    - by agazerboy
    Hi All, I asked a question before but duffymo said it is not clear so i am going to post it again here. I am using Jama api for SVD calculation. I know very well about jama and SVD. Jama does not work if your column are more than rows. I have this situation. What should I do?? any help? I can't transpose the matrix too as it can produce wrong results. Thanks. P.S: I am calculating LSI with the help of jama. I am going like column(docs) and rows ( terms )

    Read the article

  • Calculating the null space of a matrix

    - by Ainsworth
    I'm attempting to solve a set of equations of the form Ax = 0. A is known 6x6 matrix and I've written the below code using SVD to get the vector x which works to a certain extent. The answer is approximately correct but not good enough to be useful to me, how can I improve the precision of the calculation? Lowering eps below 1.e-4 causes the function to fail. from numpy.linalg import * from numpy import * A = matrix([[0.624010149127497 ,0.020915658603923 ,0.838082638087629 ,62.0778180312547 ,-0.336 ,0], [0.669649399820597 ,0.344105317421833 ,0.0543868015800246 ,49.0194290212841 ,-0.267 ,0], [0.473153758252885 ,0.366893577716959 ,0.924972565581684 ,186.071352614705 ,-1 ,0], [0.0759305208803158 ,0.356365401030535 ,0.126682113674883 ,175.292109352674 ,0 ,-5.201], [0.91160934274653 ,0.32447818779582 ,0.741382053883291 ,0.11536775372698 ,0 ,-0.034], [0.480860406786873 ,0.903499596111067 ,0.542581424762866 ,32.782593418975 ,0 ,-1]]) def null(A, eps=1e-3): u,s,vh = svd(A,full_matrices=1,compute_uv=1) null_space = compress(s <= eps, vh, axis=0) return null_space.T NS = null(A) print "Null space equals ",NS,"\n" print dot(A,NS)

    Read the article

  • MATLAB: svds() not converging

    - by Paul
    So using MATLAB's svds() function on some input data as such: [U, S, V, flag] = svds(data, nSVDs, 'L') I noticed that from run to run with the same data, I'd get drastically different output SVD sizes from run to run. When I checked whether 'flag' was set, I found that it was, indicating that the SVDs had not converged. My normal system here would be that if it really needs to converge, I'd do something like this: flag = 1 svdOpts = struct('tol', 1e-10, 'maxit', 600, 'disp', 0); while flag: if svdOpts.maxit > 1e6 error('There''s a real problem here.') end [U, S, V, flag] = svds(data, nSVDs, 'L', svdOpts) svdOpts.maxit = svdOpts.maxit*2 end But from what I can tell, when you use 'L' as the third argument, the fourth argument is ignored, meaning I just have to deal with the fact that it's not converging? I'm not even really sure how to use the 'sigma' argument in place of the 'L' argument. I've also tried reducing the number of SVDs calculated to no avail. Any help on this matter would be much appreciated. EDIT While following up on the comments below, I found that the problem had to do with the way I was building my data matrices. Turned out I had accidentally inverted a matrix and had an input of size (4000x1) rather than (20x200), which was what was refusing to converge. I also did some more timing tets and found that the fourth argument was not, in fact, being ignored, so that's on me. Thanks for the help guys.

    Read the article

  • sql perfomance on new server

    - by Rapunzo
    My database is running on a pc (AMD Phenom x6, intel ssd disk, 8GB DDR3 RAM and windows 7 OS + sql server 2008 R2 sp3 ) and it started working hard, timeout problems and up to 30 seconds long queries after 200 mb of database And I also have an old server pc (IBM x-series 266: 72*3 15k rpm scsi discs with raid5, 4 gb ram and windows server 2003 + sql server 2008 R2 sp3 ) and same query start to give results in 100 seconds.. I tried query analyser tool for tuning my indexed. but not so much improvements. its a big dissapointment for me. because I thought even its an old server pc it should be more powerfull with 15k rpm discs with raid5. what should I do. do I need $10.000 new server to get a good performance for my sql server? cant I use that IBM server? Extra information: there is 50 sql users and its an ERP program. There is my query ALTER FUNCTION [dbo].[fnDispoTerbiye] ( ) RETURNS TABLE AS RETURN ( SELECT MD.dispoNo, SV.sevkNo, M1.musteriAdi AS musteri, SD.tipTurId, TT.tipTur, SD.tipNo, SD.desenNo, SD.varyantNo, SUM(T.topMetre) AS toplamSevkMetre, MD.dispoMetresi, DT.gelisMetresi, ISNULL(DT.fire, 0) AS fire, SV.sevkTarihi, DT.gelisTarihi, SP.mamulTermin, SD.miktar AS siparisMiktari, M.musteriAdi AS boyahane, MD.akisNotu AS islemler, --dbo.fnAkisIslemleri(MD.dispoNo) DT.partiNo, DT.iplikBoyaId, B.tanimAd AS BoyaTuru, MAX(HD.hamEn) AS hamEn, MAX(HD.hamGramaj) AS hamGramaj, TS.mamulEn, TS.mamulGramaj, DT.atkiCekmesi, DT.cozguCekmesi, DT.fiyat, DV.dovizCins, DT.dovizId, (SELECT CASE WHEN DT.dovizId = 2 THEN CAST(round(SUM(T .topMetre) * DT.fiyat * (SELECT TOP 1 satis FROM tblKur WHERE dovizId = 2 ORDER BY tarih DESC), 2) AS numeric(18, 2)) WHEN DT.dovizId = 3 THEN CAST(round(SUM(T .topMetre) * DT.fiyat * (SELECT TOP 1 satis FROM tblKur WHERE dovizId = 3 ORDER BY tarih DESC), 2) AS numeric(18, 2)) WHEN DT.dovizId = 1 THEN CAST(round(SUM(T .topMetre) * DT.fiyat * (SELECT TOP 1 satis FROM tblKur WHERE dovizId = 1 ORDER BY tarih DESC), 2) AS numeric(18, 2)) END AS Expr1) AS ToplamTLfiyat, DT.aciklama, MD.dispoNotu, SD.siparisId, SD.siparisDetayId, DT.sqlUserName, DT.kayitTarihi, O.orguAd, 'Çözgü=(' + (SELECT dbo.fnTipIplikler(SD.tipTurId, SD.tipNo, SD.desenNo, SD.varyantNo, 1) AS Expr1) + ')' + ' Atki=(' + (SELECT dbo.fnTipIplikler(SD.tipTurId, SD.tipNo, SD.desenNo, SD.varyantNo, 2) AS Expr1) + ')' AS iplikAciklama, DT.prosesOk, dbo.[fnYikamaTalimat](SP.siparisId) yikamaTalimati FROM tblDoviz AS DV WITH(NOLOCK) INNER JOIN tblDispoTerbiye AS DT WITH(NOLOCK) INNER JOIN tblTanimlar AS B WITH(NOLOCK) ON DT.iplikBoyaId = B.tanimId AND B.tanimTurId = 2 ON DV.id = DT.dovizId RIGHT OUTER JOIN tblMusteri AS M1 WITH(NOLOCK) INNER JOIN tblSiparisDetay AS SD WITH(NOLOCK) INNER JOIN tblDispo AS MD WITH(NOLOCK) ON SD.siparisDetayId = MD.siparisDetayId INNER JOIN tblTipTur AS TT WITH(NOLOCK) ON SD.tipTurId = TT.tipTurId INNER JOIN tblSiparis AS SP WITH(NOLOCK) ON SD.siparisId = SP.siparisId ON M1.musteriNo = SP.musteriNo INNER JOIN tblTip AS TP WITH(NOLOCK) ON SD.tipTurId = TP.tipTurId AND SD.tipNo = TP.tipNo AND SD.desenNo = TP.desen AND SD.varyantNo = TP.varyant INNER JOIN tblOrgu AS O WITH(NOLOCK) ON TP.orguId = O.orguId INNER JOIN tblMusteri AS M WITH(NOLOCK) INNER JOIN tblSevkiyat AS SV WITH(NOLOCK) ON M.musteriNo = SV.musteriNo INNER JOIN tblSevkDetay AS SVD WITH(NOLOCK) ON SV.sevkNo = SVD.sevkNo ON MD.mamulDispoHamSevkno = SV.sevkNo LEFT OUTER JOIN tblTop AS T WITH(NOLOCK) INNER JOIN tblDispo AS HD WITH(NOLOCK) ON T.dispoNo = HD.dispoNo AND T.dispoTuruId = HD.dispoTuruId ON SVD.dispoTuruId = T.dispoTuruId AND SVD.dispoNo = T.dispoNo AND SVD.topNo = T.topNo AND MD.siparisDetayId = HD.siparisDetayId ON DT.dispoTuruId = MD.dispoTuruId AND DT.dispoNo = MD.dispoNo LEFT OUTER JOIN tblDispoTerbiyeTest AS TS WITH(NOLOCK) ON DT.dispoTuruId = TS.dispoTuruId AND DT.dispoNo = TS.dispoNo --WHERE DT.gelisTarihi IS NULL -- OR DT.gelisTarihi > GETDATE()-30 GROUP BY MD.dispoNo, DT.partiNo, DT.iplikBoyaId, TS.mamulEn, TS.mamulGramaj, DT.gelisMetresi, DT.gelisTarihi, DT.atkiCekmesi, DT.cozguCekmesi, DT.fire, DT.fiyat, DT.aciklama, DT.sqlUserName, DT.kayitTarihi, SD.tipTurId, TT.tipTur, SD.tipNo, SD.desenNo, SD.varyantNo, SD.siparisId, SD.siparisDetayId, B.tanimAd, M.musteriAdi, M.musteriAdi, M1.musteriAdi, O.orguAd, TP.iplikAciklama, SD.miktar, MD.dispoNotu, SP.mamulTermin, DT.dovizId, DV.dovizCins, MD.dispoMetresi, MD.akisNotu, SV.sevkNo, SV.sevkTarihi, DT.prosesOk,SP.siparisId )

    Read the article

  • Colloborative filtering

    - by Pranay Kumar
    How can i use SVD algorithm in mahout for producing recommendations on explicit binary data-set (eg. a user purchased or not but no specific ratings ) in an e-commerce domain ? Also what algorithms aim at producing recommendations on such binary data-sets ? Thanks in advance. Pranay Kumar, 2nd yr,cse

    Read the article

  • Efficient algorithm for finding largest eigenpair of small general complex matrix

    - by mklassen
    I am looking for an efficient algorithm to find the largest eigenpair of a small, general (non-square, non-sparse, non-symmetric), complex matrix, A, of size m x n. By small I mean m and n is typically between 4 and 64 and usually around 16, but with m not equal to n. This problem is straight forward to solve with the general LAPACK SVD algorithms, i.e. gesvd or gesdd. However, as I am solving millions of these problems and only require the largest eigenpair, I am looking for a more efficient algorithm. Additionally, in my application the eigenvectors will generally be similar for all cases. This lead me to investigate Arnoldi iteration based methods, but I have neither found a good library nor algorithm that applies to my small general complex matrix. Is there an appropriate algorithm and/or library?

    Read the article

  • Official definition of CSCI (Computer Software Configuration Item)

    - by Andreas_D
    I'm looking for the most official definition of CSCI / Configuration Item - not just what it is but what we have to deliver / can expect when a contract defines subsystems which shall be developed as configuration items. I spend some time with my famous search tool and found a lot of explanations for CSCI (wikipedia, acronym directories, ...) but I haven't found a standard or a pointer to a standard (like ISO-xxx) yet which tells (1) what it is and (2) what has to be done from a QM/CM point of view. I just ask, because a contractors QM representative stated during an acceptance test, that CI only requires to not forget the CI in the configuration plan and to assign a serial number ... I expected to see some SRS, SDD, ICD, SVD, SIP, ... documents and acceptance test documentation for those subsystems...

    Read the article

  • Triangulation & Direct linear transform

    - by srand
    Following Hartley/Zisserman's Multiview Geometery, Algorithm 12: The optimal triangulation method (p318), I got the corresponding image points xhat1 and xhat2 (step 10). In step 11, one needs to compute the 3D point Xhat. One such method is Direct Linear Transform (DLT), mentioned in 12.2 (p312) and 4.1 (p88). The homogenous method (DLT), p312-313, states that it finds a solution as the unit singular vector corresponding to the smallest singular value of A, thus, A = [xhat1(1) * P1(3,:)' - P1(1,:)' ; xhat1(2) * P1(3,:)' - P1(2,:)' ; xhat2(1) * P2(3,:)' - P2(1,:)' ; xhat2(2) * P2(3,:)' - P2(2,:)' ]; [Ua Ea Va] = svd(A); Xhat = Va(:,end); plot3(Xhat(1),Xhat(2),Xhat(3), 'r.'); However, A is a 16x1 matrix, resulting in a Va that is 1x1. What am I doing wrong (and a fix) in getting the 3D point? For what its worth sample data: xhat1 = 1.0e+009 * 4.9973 -0.2024 0.0027 xhat2 = 1.0e+011 * 2.0729 2.6624 0.0098 P1 = 699.6674 0 392.1170 0 0 701.6136 304.0275 0 0 0 1.0000 0 P2 = 1.0e+003 * -0.7845 0.0508 -0.1592 1.8619 -0.1379 0.7338 0.1649 0.6825 -0.0006 0.0001 0.0008 0.0010 A = <- my computation 1.0e+011 * -0.0000 0 0.0500 0 0 -0.0000 -0.0020 0 -1.3369 0.2563 1.5634 2.0729 -1.7170 0.3292 2.0079 2.6624

    Read the article

  • Incremental PCA

    - by smichak
    Hi, Lately, I've been looking into an implementation of an incremental PCA algorithm in python - I couldn't find something that would meet my needs so I did some reading and implemented an algorithm I found in some paper. Here is the module's code - the relevant paper on which it is based is mentioned in the module's documentation. I would appreciate any feedback from people who are interested in this. Micha #!/usr/bin/env python """ Incremental PCA calculation module. Based on P.Hall, D. Marshall and R. Martin "Incremental Eigenalysis for Classification" which appeared in British Machine Vision Conference, volume 1, pages 286-295, September 1998. Principal components are updated sequentially as new observations are introduced. Each new observation (x) is projected on the eigenspace spanned by the current principal components (U) and the residual vector (r = x - U(U.T*x)) is used as a new principal component (U' = [U r]). The new principal components are then rotated by a rotation matrix (R) whose columns are the eigenvectors of the transformed covariance matrix (D=U'.T*C*U) to yield p + 1 principal components. From those, only the first p are selected. """ __author__ = "Micha Kalfon" import numpy as np _ZERO_THRESHOLD = 1e-9 # Everything below this is zero class IPCA(object): """Incremental PCA calculation object. General Parameters: m - Number of variables per observation n - Number of observations p - Dimension to which the data should be reduced """ def __init__(self, m, p): """Creates an incremental PCA object for m-dimensional observations in order to reduce them to a p-dimensional subspace. @param m: Number of variables per observation. @param p: Number of principle components. @return: An IPCA object. """ self._m = float(m) self._n = 0.0 self._p = float(p) self._mean = np.matrix(np.zeros((m , 1), dtype=np.float64)) self._covariance = np.matrix(np.zeros((m, m), dtype=np.float64)) self._eigenvectors = np.matrix(np.zeros((m, p), dtype=np.float64)) self._eigenvalues = np.matrix(np.zeros((1, p), dtype=np.float64)) def update(self, x): """Updates with a new observation vector x. @param x: Next observation as a column vector (m x 1). """ m = self._m n = self._n p = self._p mean = self._mean C = self._covariance U = self._eigenvectors E = self._eigenvalues if type(x) is not np.matrix or x.shape != (m, 1): raise TypeError('Input is not a matrix (%d, 1)' % int(m)) # Update covariance matrix and mean vector and centralize input around # new mean oldmean = mean mean = (n*mean + x) / (n + 1.0) C = (n*C + x*x.T + n*oldmean*oldmean.T - (n+1)*mean*mean.T) / (n + 1.0) x -= mean # Project new input on current p-dimensional subspace and calculate # the normalized residual vector g = U.T*x r = x - (U*g) r = (r / np.linalg.norm(r)) if not _is_zero(r) else np.zeros_like(r) # Extend the transformation matrix with the residual vector and find # the rotation matrix by solving the eigenproblem DR=RE U = np.concatenate((U, r), 1) D = U.T*C*U (E, R) = np.linalg.eigh(D) # Sort eigenvalues and eigenvectors from largest to smallest to get the # rotation matrix R sorter = list(reversed(E.argsort(0))) E = E[sorter] R = R[:,sorter] # Apply the rotation matrix U = U*R # Select only p largest eigenvectors and values and update state self._n += 1.0 self._mean = mean self._covariance = C self._eigenvectors = U[:, 0:p] self._eigenvalues = E[0:p] @property def components(self): """Returns a matrix with the current principal components as columns. """ return self._eigenvectors @property def variances(self): """Returns a list with the appropriate variance along each principal component. """ return self._eigenvalues def _is_zero(x): """Return a boolean indicating whether the given vector is a zero vector up to a threshold. """ return np.fabs(x).min() < _ZERO_THRESHOLD if __name__ == '__main__': import sys def pca_svd(X): X = X - X.mean(0).repeat(X.shape[0], 0) [_, _, V] = np.linalg.svd(X) return V N = 1000 obs = np.matrix([np.random.normal(size=10) for _ in xrange(N)]) V = pca_svd(obs) print V[0:2] pca = IPCA(obs.shape[1], 2) for i in xrange(obs.shape[0]): x = obs[i,:].transpose() pca.update(x) U = pca.components print U

    Read the article

1