MATLAB Central Newsreader - index array with variable number of dimensions

index array with variable number of dimensions

scott — Mon, 10 Nov 2008 19:58:02 -0500

I have a function where I allocate a multidimensional numeric array. The array dimensionality depends upon arguments passed to the function. I'd like to know if there is a way to flexibly index the array.

For example, let's say I have a time series of 100 timepoints repeated 30 times. I want to vary two different parameters across 4 and 5 values and calculate some statistic over the repeats for each parameter combination and timepoint.

So my data could be a big array of size 100 x 30 x 4 x 5. And my statistic array would be an array of size 100 x 4 x 5.

That's no problem. But what if I want to make my function flexible so that if I wanted to vary 3 parameters instead of 2 I could do it with the same code. So let's say I have another parameter that I vary over 6 values. Then my data array would be size 100 x 30 x 4 x 5 x 6, and my statistic array would be size 100 x 4 x 5 x 6.

My problem is that I don't see any way to do the indexing. If I want to assign a value to the statistic array, I can do it if I know how many dimensions there are: statisticArray(:,2,3) = mean(dataArray(:,:,2,3),2) but if there is sometimes an extra dimension then it would have to be statisticArray(:,2,3,4)=mean(dataArray(:,:,2,3,4)).

Is there any way to have a variable number of dimensions and, say, put the indexes I want for these into a 1D vector and then just index something like: statisticArray(:,[2 3 4])??

(note that the last thing doesn't work, but it gives the idea of what I'd like)

Thanks much.

Re: index array with variable number of dimensions

Loren Shure — Mon, 10 Nov 2008 20:32:18 -0500

In article <gfa3oa$ej8$1@fred.mathworks.com>,
grenblau@yahoo.deletethis.com says...
> I have a function where I allocate a multidimensional numeric array. The array dimensionality depends upon arguments passed to the function. I'd like to know if there is a way to flexibly index the array.
>
> For example, let's say I have a time series of 100 timepoints repeated 30 times. I want to vary two different parameters across 4 and 5 values and calculate some statistic over the repeats for each parameter combination and timepoint.
>
> So my data could be a big array of size 100 x 30 x 4 x 5. And my statistic array would be an array of size 100 x 4 x 5.
>
> That's no problem. But what if I want to make my function flexible so that if I wanted to vary 3 parameters instead of 2 I could do it with the same code. So let's say I have another parameter that I vary over 6 values. Then my data array would be size 100 x 30 x 4 x 5 x 6, and my statistic array would be size 100 x 4 x 5 x 6.
>
> My problem is that I don't see any way to do the indexing. If I want to assign a value to the statistic array, I can do it if I know how many dimensions there are: statisticArray(:,2,3) = mean(dataArray(:,:,2,3),2) but if there is sometimes an extra dimension then it would have to be statisticArray(:,2,3,4)=mean(dataArray(:,:,2,3,4)).
>
> Is there any way to have a variable number of dimensions and, say, put the indexes I want for these into a 1D vector and then just index something like: statisticArray(:,[2 3 4])??
>
> (note that the last thing doesn't work, but it gives the idea of what I'd like)
>
> Thanks much.
>

nd = ndims(x);
index = repmat({':'},1,nd);

x(index{:}) gives you nd indexing. Now just replace whichever dimension
you want in the vector index and you should be all set.

--
Loren
http://blogs.mathworks.com/loren

Re: index array with variable number of dimensions

Walter Roberson — Mon, 10 Nov 2008 20:34:19 -0500

scott wrote:

> Is there any way to have a variable number of dimensions and, say, put the indexes
> I want for these into a 1D vector and then just index something like:
> statisticArray(:,[2 3 4])??

Use a cell array.

TheseIndices = {2 3 4};
statisticArray(:,TheseIndices{:});

(And if you prefer to work with an array of indices, remember num2cell can
transform an array of numbers into a cell array of individual numbers.)

--
.signature note: I am now avoiding replying to unclear or ambiguous postings.
Please review questions before posting them. Be specific. Use examples of what you mean,
of what you don't mean. Specify boundary conditions, and data classes and value
relationships -- what if we scrambled your data or used -Inf, NaN, or complex(rand,rand)?

Re: index array with variable number of dimensions

scott — Mon, 10 Nov 2008 21:54:02 -0500

Walter Roberson <roberson@hushmail.com> wrote in message <YU0Sk.2513$lo6.2424@newsfe01.iad>...
> scott wrote:
>
> > Is there any way to have a variable number of dimensions and, say, put the indexes
> > I want for these into a 1D vector and then just index something like:
> > statisticArray(:,[2 3 4])??
>
> Use a cell array.
>
> TheseIndices = {2 3 4};
> statisticArray(:,TheseIndices{:});
>
> (And if you prefer to work with an array of indices, remember num2cell can
> transform an array of numbers into a cell array of individual numbers.)
>
>
> --

That seems to have done it - thanks Walter and Loren

Re: index array with variable number of dimensions

Walter Roberson — Mon, 10 Nov 2008 22:46:09 -0500

Loren Shure wrote:

> nd = ndims(x);
> index = repmat({':'},1,nd);

> x(index{:}) gives you nd indexing.

Interesting -- you can't use 'end' though, and you have to use numeric cell
arrays instead of strings of numbers... e.g.,

index = {':' 9 ':'};
x(index{:})

-not-

index = {':' '9' ':'};
x(index{:})

The ':' must be treated as a special case ?

--
.signature note: I am now avoiding replying to unclear or ambiguous postings.
Please review questions before posting them. Be specific. Use examples of what you mean,
of what you don't mean. Specify boundary conditions, and data classes and value
relationships -- what if we scrambled your data or used -Inf, NaN, or complex(rand,rand)?

Re: index array with variable number of dimensions

Loren Shure — Tue, 11 Nov 2008 14:00:52 -0500

In article <yQ2Sk.3011$Gm3.2484@newsfe01.iad>, roberson@hushmail.com
says...
> Loren Shure wrote:
>
> > nd = ndims(x);
> > index = repmat({':'},1,nd);
>
> > x(index{:}) gives you nd indexing.
>
> Interesting -- you can't use 'end' though, and you have to use numeric cell
> arrays instead of strings of numbers... e.g.,
>
> index = {':' 9 ':'};
> x(index{:})
>
> -not-
>
> index = {':' '9' ':'};
> x(index{:})
>
> The ':' must be treated as a special case ?
>
>

yes, and if you create and use the cell area inside an indexed
expression, then you can use end, (not 'end').

x(max(10,end),1)
--
Loren
http://blogs.mathworks.com/loren

Re: index array with variable number of dimensions

Walter Roberson — Mon, 10 Nov 2008 20:15:50 -0500