How to combine cellfun and strfind?
in [Matlab]
|
Prev: b2bsharing
Next: Dynamic quantile regression
From: Emil on 27 Jul 2010 03:40 The solution proposed by us str = '001'; c={'.','..','ab001','ab002','ac001'}; ix=~cellfun(@isempty,regexp(c,str)) I find very elegant. Unfortunately does it only work on char arrays. Can this be modified such that it works for cell arrays of strings as well? The output then would be a cell array with the same amount of elements as XX. So the statement I am looking for should behave like str = {'001','b0'}; c={'.','..','ab001','ab002','ac001'}; ix=~cellfun(@isempty,regexp(c,str)) ix{1} = [0 0 1 0 0]; ix{2} = [0 0 1 1 0]; Any ideas?
From: us on 27 Jul 2010 06:05 "Emil " <emil4ml(a)gmail.com> wrote in message <i2m2go$ie2$1(a)fred.mathworks.com>... > The solution proposed by us > str = '001'; > c={'.','..','ab001','ab002','ac001'}; > ix=~cellfun(@isempty,regexp(c,str)) > I find very elegant. Unfortunately does it only work on char arrays. Can this be modified such that it works for cell arrays of strings as well? > > The output then would be a cell array with the same amount of elements as XX. So the statement I am looking for should behave like > str = {'001','b0'}; > c={'.','..','ab001','ab002','ac001'}; > ix=~cellfun(@isempty,regexp(c,str)) > > ix{1} = [0 0 1 0 0]; > ix{2} = [0 0 1 1 0]; > > Any ideas? one of the (more contrived) solutions str={'001','b0'}; c={'.','..','ab001','ab002','ac001'}; ix=arrayfun(@(x) ~cellfun(@isempty,regexp(c,str(x))),1:numel(str),'uni',false); ix=cat(1,ix{:}) %{ % ix = 0 0 1 0 1 0 0 1 1 0 %} us
From: Emil on 28 Jul 2010 02:52 Thanks, with the help your post in the other thread I came up with the same solution. It looks really elegant! However, when I compared the performance to that of a for-loop I found tic for i = 1:n for j = 1:numel(id) ix{j} = ~cellfun(@isempty,regexp(c,str{j})); end end toc to be 2.5 times faster. I went on to check other examples and found arrayfun to always be slower than the for-loop it is replacing. When is it, performance-wise, an advantage to use arrayfun?
From: Oleg Komarov on 28 Jul 2010 03:41 "Emil " <emil4ml(a)gmail.com> wrote in message <i2ok2m$75q$1(a)fred.mathworks.com>... > Thanks, with the help your post in the other thread I came up with the same solution. It looks really elegant! However, when I compared the performance to that of a for-loop I found > > tic > for i = 1:n > for j = 1:numel(id) > ix{j} = ~cellfun(@isempty,regexp(c,str{j})); > end > end > toc > > to be 2.5 times faster. I went on to check other examples and found arrayfun to always be slower than the for-loop it is replacing. When is it, performance-wise, an advantage to use arrayfun? Never. Only cellfun has 3 syntaxes (the last 3 ones listed in the help) which may be at least as fast as a loop. Oleg
From: Matt Fig on 28 Jul 2010 11:21
"Emil " <emil4ml(a)gmail.com> wrote in message <i2ok2m$75q$1(a)fred.mathworks.com>... > Thanks, with the help your post in the other thread I came up with the same solution. It looks really elegant! However, when I compared the performance to that of a for-loop I found > > tic > for i = 1:n > for j = 1:numel(id) > ix{j} = ~cellfun(@isempty,regexp(c,str{j})); > end > end > toc > For even more speed use 'isempty' instead of @isempty. |