Sparse matrix with determined number of zeros
in [Matlab]
|
Prev: how to make embedded matlab compliant and remove "non-scalars of class 'function_handle is unsupported","CAT arguments dimension are not consistent|
Next: Surface generation
From: us on 23 Jun 2010 06:55 "Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message <hvso4g$589$1(a)fred.mathworks.com>... > "us " <us(a)neurol.unizh.ch> wrote in message <hvsnjk$1ju$1(a)fred.mathworks.com>... > > haithem <haithem.haggui(a)gmail.com> wrote in message <ae397350-a6bc-4935-8d00-3beb656223b0(a)g19g2000yqc.googlegroups.com>... > > > Hi, all. > > > > > > My problem is that I want to generate a sparse vector with a fixed > > > number of non zero elements randomly distributed. > > > > > > Thanks for all > > > > one of the many (other) solutions > > > > n=10; > > nr=2; > > s=spalloc(1,10,10); > > rp=randperm(n); > > s(rp(1:nr))=1; > > > > us > > The problem is for large sparse matrix size such as m x n = 1e6 x 1e6, generate a randperm of size 1e12 is a no-no. > > Bruno ok... a pedestrian work-around might look like this ns=10; % <- an arbitrary value >> 1... rp=unique(ceil(n*rand(1,ns*nr))); % <- changes are slim they are all equal... s(rp(1:nr))=1; us
From: Bruno Luong on 23 Jun 2010 07:10 "us " <us(a)neurol.unizh.ch> wrote in message <hvsp6q$d4k$1(a)fred.mathworks.com>... > "Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message <hvso4g$589$1(a)fred.mathworks.com>... > > "us " <us(a)neurol.unizh.ch> wrote in message <hvsnjk$1ju$1(a)fred.mathworks.com>... > > > haithem <haithem.haggui(a)gmail.com> wrote in message <ae397350-a6bc-4935-8d00-3beb656223b0(a)g19g2000yqc.googlegroups.com>... > > > > Hi, all. > > > > > > > > My problem is that I want to generate a sparse vector with a fixed > > > > number of non zero elements randomly distributed. > > > > > > > > Thanks for all > > > > > > one of the many (other) solutions > > > > > > n=10; > > > nr=2; > > > s=spalloc(1,10,10); > > > rp=randperm(n); > > > s(rp(1:nr))=1; > > > > > > us > > > > The problem is for large sparse matrix size such as m x n = 1e6 x 1e6, generate a randperm of size 1e12 is a no-no. > > > > Bruno > > ok... > a pedestrian work-around might look like this > > ns=10; % <- an arbitrary value >> 1... > rp=unique(ceil(n*rand(1,ns*nr))); % <- changes are slim they are all equal... > s(rp(1:nr))=1; This is almost SPRAND does. Changes are indeed slim unfortunately slim is not nil. We could over fill then trim down as John has suggested, but strim down is also costly: m=1e6; n=1e6; nz=1e6; S = sparse([], [], [], m, n); while nnz(S) < nz p = round(1.1*nz) - nnz(S); % 10 % overfilled i = ceil(m*rand(p,1)); j = ceil(n*rand(p,1)); S = S + sparse(i, j, 1, m, n); end if nnz(S) > nz [i j] = find(S); p = randperm(length(i)); i = i(p(1:nz)); j = j(p(1:nz)); S = sparse(i, j, 1, m, n); end Bruno
From: us on 23 Jun 2010 07:35 "Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message <hvsq2g$8or$1(a)fred.mathworks.com>... > "us " <us(a)neurol.unizh.ch> wrote in message <hvsp6q$d4k$1(a)fred.mathworks.com>... > > "Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message <hvso4g$589$1(a)fred.mathworks.com>... > > > "us " <us(a)neurol.unizh.ch> wrote in message <hvsnjk$1ju$1(a)fred.mathworks.com>... > > > > haithem <haithem.haggui(a)gmail.com> wrote in message <ae397350-a6bc-4935-8d00-3beb656223b0(a)g19g2000yqc.googlegroups.com>... > > > > > Hi, all. > > > > > > > > > > My problem is that I want to generate a sparse vector with a fixed > > > > > number of non zero elements randomly distributed. > > > > > > > > > > Thanks for all > > > > > > > > one of the many (other) solutions > > > > > > > > n=10; > > > > nr=2; > > > > s=spalloc(1,10,10); > > > > rp=randperm(n); > > > > s(rp(1:nr))=1; > > > > > > > > us > > > > > > The problem is for large sparse matrix size such as m x n = 1e6 x 1e6, generate a randperm of size 1e12 is a no-no. > > > > > > Bruno > > > > ok... > > a pedestrian work-around might look like this > > > > ns=10; % <- an arbitrary value >> 1... > > rp=unique(ceil(n*rand(1,ns*nr))); % <- changes are slim they are all equal... > > s(rp(1:nr))=1; > > This is almost SPRAND does. Changes are indeed slim unfortunately slim is not nil. of course... however, i wonder - for practical reasons - how slimly slim approaches nil; i guess: nil... us
From: Bruno Luong on 23 Jun 2010 08:10 "us " <us(a)neurol.unizh.ch> wrote in message <hvsrh8$c47$1(a)fred.mathworks.com>... > "Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message > > of course... > however, i wonder - for practical reasons - how slimly slim approaches nil; i guess: nil... > If I might give a suggestion us, when talking to women please don't mix between "you are slim" and "you are nil". Bruno
From: haithem on 5 Jul 2010 15:07
On 23 juin, 06:28, "us " <u...(a)neurol.unizh.ch> wrote: > haithem <haithem.hag...(a)gmail.com> wrote in message <ae397350-a6bc-4935-8d00-3beb65622...(a)g19g2000yqc.googlegroups.com>... > > Hi, all. > > > My problem is that I want to generate a sparse vector with a fixed > > number of non zero elements randomly distributed. > > > Thanks for all > > one of the many (other) solutions > > n=10; > nr=2; > s=spalloc(1,10,10); > rp=randperm(n); > s(rp(1:nr))=1; > > us It's exactly what I need!! Many thanks |