Comments on: Scalar Expansion and More http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/ Loren Shure works on design of the MATLAB language at <a href="http://www.mathworks.com/">The MathWorks</a>. She writes here about once a week on MATLAB programming and related topics. <br><br><a href="/images/loren-full.jpg"><img src="/images/loren.jpg"></a> Mon, 23 Nov 2009 01:09:55 +0000 http://wordpress.org/?v=2.3.1 By: StephenLL http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-30271 StephenLL Tue, 05 May 2009 15:04:55 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-30271 You can also check out my _apply_ function at the FEX. http://www.mathworks.com/matlabcentral/fileexchange/16467 While it doesn't speed anything up, it makes the code somewhat cleaner. I wish the mathworks would add something like this to base MATLAB. Stephen You can also check out my _apply_ function at the FEX.

http://www.mathworks.com/matlabcentral/fileexchange/16467

While it doesn’t speed anything up, it makes the code somewhat cleaner. I wish the mathworks would add something like this to base MATLAB.

Stephen

]]> By: Loren http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-30269 Loren Mon, 04 May 2009 11:23:57 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-30269 Everett- Many MATLAB functions already work on columns of an array or can work on a selected dimension (see help for max, for example). --Loren Everett-

Many MATLAB functions already work on columns of an array or can work on a selected dimension (see help for max, for example).

–Loren

]]> By: Everett http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-30267 Everett Sun, 03 May 2009 15:17:29 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-30267 Hi, Loren, How to apply a function to each row of a matrix with out using the for loop? Thanks, Everett Hi, Loren,

How to apply a function to each row of a matrix with out using the for loop?

Thanks,
Everett

]]> By: Loren http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-29873 Loren Tue, 18 Nov 2008 18:08:49 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-29873 Tom- See the function <a href="http://www.mathworks.com/access/helpdesk/help/techdoc/ref/bsxfun.html" rel="nofollow">bsxfun</a> for other scalar expansion functionality. --Loren Tom-

See the function bsxfun for other scalar expansion functionality.

–Loren

]]> By: Tom Clark http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-29872 Tom Clark Tue, 18 Nov 2008 18:04:59 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-29872 Loren, Ah ha! I can't believe I've been using MATLAB in anger, daily, for years now - and hadn't figured out the outer product. God bless you, this just saved me 586,000 seconds... And then the rest! Tom Loren,

Ah ha! I can’t believe I’ve been using MATLAB in anger, daily, for years now - and hadn’t figured out the outer product.

God bless you, this just saved me 586,000 seconds… And then the rest!

Tom

]]> By: Melchi http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-28225 Melchi Wed, 09 Apr 2008 23:36:19 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-28225 Regarding "How to Add [or multiply] Constant Values to Each Column of a Matrix." This bothered me for a long time. I used to use what Loren calls "Tony's Method" (both versions) for a while but got frustrated with it because of the performance hit for large matrices. Though I found the repmat method to be faster than looping, I couldn't get over the fact that I was doing so many extra additions (or multiplications). Finally, I wrote a few mex functions (e.g., rowmult(cVec,Matrix), colmult(rVec,Matrix), etc.) that calculate these directly. These are really basic operations that I use so often that I'm really shocked that they aren't built in to Matlab. While it might be nice to eventually come up with some operators to represent these functions, I think it's more important that efficient functions implementing these operations be included even if only in functional form. Regarding “How to Add [or multiply] Constant Values to Each Column of a Matrix.” This bothered me for a long time. I used to use what Loren calls “Tony’s Method” (both versions) for a while but got frustrated with it because of the performance hit for large matrices. Though I found the repmat method to be faster than looping, I couldn’t get over the fact that I was doing so many extra additions (or multiplications). Finally, I wrote a few mex functions (e.g., rowmult(cVec,Matrix), colmult(rVec,Matrix), etc.) that calculate these directly. These are really basic operations that I use so often that I’m really shocked that they aren’t built in to Matlab. While it might be nice to eventually come up with some operators to represent these functions, I think it’s more important that efficient functions implementing these operations be included even if only in functional form.

]]> By: Loren http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-25295 Loren Thu, 24 Jan 2008 12:22:25 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-25295 Peter- I don't know what you are trying to calculate. Are the columns year, month, and number. You'd probably be best off reshaping your array into a 3-dimensional array and summing along the correct dimension(s) then. --Loren Peter-

I don’t know what you are trying to calculate. Are the columns year, month, and number. You’d probably be best off reshaping your array into a 3-dimensional array and summing along the correct dimension(s) then.

–Loren

]]> By: Peter Maclean http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-25271 Peter Maclean Wed, 23 Jan 2008 22:10:16 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-25271 How to calculate a cummulative sum of a matrix over over a given varible. Example, how I calculate year totals from monthly total. The data set look like Country month Temp 1 1 10 . . . 1 12 12 2 1 60 . . . 2 12 72 How to calculate a cummulative sum of a matrix over over a given varible. Example, how I calculate year totals from monthly total. The data set look like

Country month Temp
1 1 10
. . .
1 12 12
2 1 60
. . .
2 12 72

]]> By: Lee Newman http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-16444 Lee Newman Fri, 31 Aug 2007 04:07:09 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-16444 I've recently switched from Mathematica to Matlab as my primary programming tool, primarily because Matlab was orders of magnitude faster for my purposes (self organizing neural networks). However, I do think that Matlab developers ought to consider the way Mathematica "vectorizes". Almost every function when given a N-D matrix performs its operations row-wise and via the various "functional programming" functions (Map, MapAt, MapThread, Apply, etc) which allow the coder to control exactly how, and at what level, functions operate on data structures. I've found it takes many more lines of code to vectorize in Matlab. Perhaps it's due to my relative lack of familiarity with Matlab, but I do think Matlab might benefit by incorporating some of the vectorization methods from Mathematica. I’ve recently switched from Mathematica to Matlab as my primary programming tool, primarily because Matlab was orders of magnitude faster for my purposes (self organizing neural networks). However, I do think that Matlab developers ought to consider the way Mathematica “vectorizes”. Almost every function when given a N-D matrix performs its operations row-wise and via the various “functional programming” functions (Map, MapAt, MapThread, Apply, etc) which allow the coder to control exactly how, and at what level, functions operate on data structures. I’ve found it takes many more lines of code to vectorize in Matlab. Perhaps it’s due to my relative lack of familiarity with Matlab, but I do think Matlab might benefit by incorporating some of the vectorization methods from Mathematica.

]]> By: Stephen http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-78 Stephen Tue, 14 Mar 2006 15:02:11 +0000 http://blogs.mathworks.com/loren/2006/02/22/scalar-expansion-and-more-take-2/#comment-78 I did a quick comparison of Doug's matlab only code vs c code versions of the genops for @plus and I'm seeing the matlab code version take 35 times longer to run then the c code version. I appreciate having both around, but I'd like to have function handles be passed to the c version. Unfortunately I know making genop have the same flexibility of cellfun, arrayfun, etc... is well beyond what I can handle in mex functions. If the source code for the ones already in matlab were available maybe we can make it a community project to create a gopfun mex function. Stephen I did a quick comparison of Doug’s matlab only code vs c code versions of the genops for @plus and I’m seeing the matlab code version take 35 times longer to run then the c code version.

I appreciate having both around, but I’d like to have function handles be passed to the c version. Unfortunately I know making genop have the same flexibility of cellfun, arrayfun, etc… is well beyond what I can handle in mex functions. If the source code for the ones already in matlab were available maybe we can make it a community project to create a gopfun mex function.

Stephen

]]>