spfun
in [Matlab]
|
From: Bruno Luong on 23 Nov 2009 13:21 "Matt " <xys(a)whatever.com> wrote in message <heeinc$ha7$1(a)fred.mathworks.com>... > "Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message <heefss$lcl$1(a)fred.mathworks.com>... > > > %%%%%%%%%%%%%%%%%% > > > > This is one of the reason why I wrote this package > > http://www.mathworks.com/matlabcentral/fileexchange/23488-sparse-sub-access > > > > >> [i j]=find(A); > > >> AB = setsparse(A,i,j,getsparse(B,i,j),@(a,b) a./b) > > > > AB = > > > > (100000,100000) 1 > > > > % Bruno > > > I didn't realize the extent of the problem. It wonder if one could provide a subclass of @double to interface to this package. That way, we could use the old indexing syntax instead of things like this Could be a nice project indeed, if there is demand out there. I'm not going to involve in coding it though (lack of time). Bruno
From: Matt on 23 Nov 2009 14:38 "Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message <heejqv$q09$1(a)fred.mathworks.com>... > > Could be a nice project indeed, if there is demand out there. I'm not going to involve in coding it though (lack of time). > ======================= Would you be willing to at least test-drive the class I've coded at the bottom of this post (to be put in a file called RobustSparse.m). It will take me some time to fully understand/install your package, whereas my code would be a 1-file installation for you. If you're willing, the following revised tests should give the correct results. Otherwise, it will default with a warning to the normal sparse indexing behavior >> A=RobustSparse(100000,100000,1), B=A B = (100000,100000) 1 >> result=A; >> idx=logical(A); >> result(idx)=A(idx)./B(idx) %%%%Put in file RobustSparse.m classdef RobustSparse < double %RobustSparse - a class essentially the same as sparse @double, but with %some bug protection in subsref/subsasgn courtesy of Bruno Luong. methods function obj=RobustSparse(varargin) %constructor for RobustSparse data=sparse(varargin{:}); obj=obj(a)double(data); end function objnew=subsref(obj,S) %SUBSREF for RobustSparse class - use Bruno Luong's routines nn=length(S.subs); idx=S.subs{1}; if nn==1 && ~ischar(idx) try if islogical(idx) [i,j]=find(idx); else%linear index [i,j]=ind2sub(size(obj),idx); end objnew=getsparse(double(obj),i,j); objnew=reshape(objnew,size(idx)); catch%Default to usual method disp 'sparse-sub-access not available. Defaulting...' objnew=subsref(a)double(obj,S); end else objnew=subsref(a)double(obj,S); end objnew=RobustSparse(objnew); end function objnew=subsasgn(obj,S,rhs) %SUBSASGN for RobustSparse class - use Bruno Luong's routines nn=length(S.subs); idx=S.subs{1}; if nn==1 && ~ischar(idx) try if islogical(idx) [i,j]=find(idx); else%single linear index [i,j]=ind2sub(size(obj),idx); end objnew=setsparse(double(obj),i,j,rhs); catch%Default to usual method disp 'sparse-sub-access not available. Defaulting...' objnew=subsasgn(a)double(obj,S,rhs); end else objnew=subsasgn(a)double(obj,S,rhs); end objnew=RobustSparse(objnew); end function obj=uplus(obj) end function obj=uminus(obj) obj=RobustSparse(uminus(a)double(obj)); end function obj=plus(L,R) obj = RobustSparse( plus(a)double(L,R) ); end function obj=minus(L,R) obj = RobustSparse( minus(a)double(L,R) ); end function obj=times(L,R) obj = RobustSparse( times(a)double(L,R) ); end function obj=mtimes(L,R) obj = RobustSparse( mtimes(a)double(L,R) ); end function obj=rdivide(L,R) obj = RobustSparse( rdivide(a)double(L,R) ); end function obj=ldivide(L,R) obj = RobustSparse( ldivide(a)double(L,R) ); end function obj=mrdivide(L,R) obj = RobustSparse( mrdivide(a)double(L,R) ); end function obj=mldivide(L,R) obj = RobustSparse( mldivide(a)double(L,R) ); end function obj=inv(L) obj = RobustSparse( inv(a)double(L) ); end function obj=sum(L,varargin) obj = RobustSparse( sum(a)double(L,varargin{:}) ); end function display(obj) l=inputname(1); if isempty(l), l='ans'; end disp ' ' disp([l ' = ']); disp ' ' disp(double(obj)); end end end
From: Bruno Luong on 23 Nov 2009 14:57 "Matt " <xys(a)whatever.com> wrote in message <heeoar$798$1(a)fred.mathworks.com>... > > If you're willing, the following revised tests should give the correct results. Otherwise, it will default with a warning to the normal sparse indexing behavior > I sure will, but I cannot take a closer look before Wednesday for many reasons. I'm sure you will make the indexing code right Matt as you seems to be very keen on oop. Thanks, Bruno
From: Matt on 23 Nov 2009 15:22 "Bruno Luong" <b.luong(a)fogale.findmycountry> wrote in message <heepeh$hq1$1(a)fred.mathworks.com>... >I'm sure you will make the indexing code right Matt as you seems to be very keen on oop. ==== Well this seems like a particularly good application of it, although hopefully TMW will fix the bug soon and render all this unnecessary.
From: arl on 26 Nov 2009 05:29 On 23 Nov, 18:02, "Matt " <x...(a)whatever.com> wrote: > "Bruno Luong" <b.lu...(a)fogale.findmycountry> wrote in message <heefss$lc....(a)fred.mathworks.com>... > > %%%%%%%%%%%%%%%%%% > > > This is one of the reason why I wrote this package Thank you for the help! I also didn't realize that problem... > >http://www.mathworks.com/matlabcentral/fileexchange/23488-sparse-sub-... > > > >> [i j]=find(A); > > >> AB = setsparse(A,i,j,getsparse(B,i,j),@(a,b) a./b) > > > AB = > > > (100000,100000) 1 > > > % Bruno > > I didn't realize the extent of the problem. It wonder if one could provide a subclass of @double to interface to this package. That way, we could use the old indexing syntax instead of things like this > > AB = setsparse(A,i,j,getsparse(B,i,j),@(a,b) a./b)
First
|
Prev
|
Next
|
Last
Pages: 1 2 3 4 Prev: PCA reconstruction error Next: How to count the number of each intensity value? |