MATLAB Basics: Nested functions

In a previous video we covered functions in MATLAB. This week we will be covering nested subfunctions. This style of subfunction allows you to share the workspace of the calling function without explicitly passing variables.

Loren did a nice job covering this topic also. [click here] As always, she is looking at more advanced topics in MATLAB than I tend to. You should stop by her main page to see what she is up to each week. [click here]

Video Content

You can follow any responses to this entry through the RSS 2.0 feed. You can skip to the end and leave a response. Pinging is currently not allowed.

11 Responses to “MATLAB Basics: Nested functions”

quan replied on February 2nd, 2008 at 10:47 pm :

Nice tutorial Doug. I’ve never used nested subfunctions, but I can see them being very useful.

Maria C Marcano replied on February 4th, 2008 at 10:01 pm :

Is there a way to suscribe to these demos as podcasts, or do I need to come and download them from the page each week?
Thanks!

Doug replied on February 5th, 2008 at 8:47 am :

In iTunes, you can subscribe to the RSS feed and it will find the podcasts that I embed into the pages.

Doug

adnyana replied on February 15th, 2008 at 5:17 am :

MR. Doug, i want ask you something. I have a problem in doing
subdivision of segmented image into piecewise of images. but this
program(code) need much time to complete computation of image 512×512
pixel(need about 3 thousand second for one iteration). here is the code
:

function hasil = Subdivision1(cls,img,img_asli,sigma)

global Wr pxl kls rw img_asl img_result cpt;

…..etc….

[Long code deleted from comment -Doug]

Mr. Doug please help me
i need your suggestion

thanks a lot
best regard

adnyana

Doug replied on February 15th, 2008 at 12:58 pm :

Adnyana,

I think that you would want to run the profiler on this code, it will tell you where the slow parts are. From there, we could figure out how to speed it up.

profiler:
http://blogs.mathworks.com/pick/2006/10/19/profiler-to-find-code-bottlenecks/

Doug

adnyana replied on February 16th, 2008 at 5:38 am :

yes sir, i’ve run profiler on this code.
but the most time was spent on the ‘ismember’ function.

is there other way or other function that run faster than ‘ismember’.

i use ‘ismember’ function to check if one set of data is member of other set, like this :

ismember(ind,kls)== 1 && ismember(ind,pxl) == 0

thanks a lot sir for helping me

best regard

adnyana

adnyana replied on February 16th, 2008 at 5:54 am :

sorry sir the image not appear, here i try again :

http://img151.imageshack.us/img151/2596/profilersp5.jpg

Doug replied on February 19th, 2008 at 8:57 am :

Adnyana,

This is great. Now, all I need to know is “What are the typical values of:

ind
kls
pxl

Are these scalars, vectors, matrices? What are the typical sizes. Do the values change often? If they do not change, can you make these calculations once and then store the value?

Doug

adnyana replied on February 21st, 2008 at 12:46 am :

ind is scalar, kls and pxl are vectors
yes of course, all of those variables’ values are change very often, as stated in my code. kls store indexes of one class of image, pxl store indexes of one region of that class.

here is segment of the code :

function IncludeInRjPixel_1(ind) %horizontal & vertical
global Wr pxl kls rw cpt;
flag = 0;
% cpt = 0;
f = 0;
while cpt index of image matrix
case 1
ind = ind + 1;
case 2
ind = ind - rw;
case 3
ind = ind + rw;
end;
if f >=4
break;
end;
end;
includePixel(pxl); % calculate average of pixel value of index stored in pxl

this code is called very often until computation is complete

so i can’t make this calculation once. so what do you think of this case sir?

thanks sir for your reply

best regard

adnyana

Doug replied on February 21st, 2008 at 3:16 pm :

Adnyana,

It looks like you are spending 22 hours on this one function, but you are calling it so often that each call is 0.001 seconds. My guess is this is actually quite efficient. Maybe your algorithm can be changed to call ismember.m less often? ismember.m has some decisions internally to decide which algorithm to use based on the lengths of the inputs. You might be able to tweak these.

My other suggestion is this looks like a problem that might call for a computer cluster. You can really cut down on the time needed by using more computing power. Check out the Parallel Computing Toolbox.

Doug

ALT replied on June 7th, 2008 at 6:53 am :

Thanks for this video it was very useful.
I was using 3 functions, using global variables, but now its just one function with two nested subfunctions.

You are doing a great job with your videos, they are very helpful especially for beginners like me.

Altino

About

Doug Hull is an Application Engineer at The MathWorks. A MATLAB user since 1994, he gets paid to live, eat, and breathe MATLAB! This blog is dedicated to promoting the File Exchange by highlighting files and original video content.

Doug’s Pick of the Week

February 1st, 2008