Find the index of cell array containing numeric values

Hi!I have a cell array=YTLim, which contains at each cell the measurements of each for significant wave height Hs(numeric values).Each cell has 2920=365*8 or 2928=366*8 measurements,it depends on the year.So first cell has the values

YTLim{1}=[1,2,3,...,2920]
YTLim{2}=[2921,2922,2923,...,5840] etc

I have,also, a vector that contains the spot of some measurements. For example:

maxpos=[242;365;...;8555]

I need to find a way to take the index of the cell containing the vector's value...Is that possible?

I need something like that:

YS=[1;1;2;...;10;11;11]

Answer 1

[the cyclist]

Here is a miniature example that I believe does what you've asked for:

YTLim{1}=[1,2,3,4];
YTLim{2}=[5,6,7,8];
YTLim{3}=[9,10,11,12,13];

maxpos=[1,3,7,13];

c = cellfun(@(x)(ismember(maxpos,x)),YTLim,'UniformOutput',false);
[YS,~] = find(reshape([c{:}],numel(maxpos),[])')

I have to admit the code is compact and maybe a bit obscure. But if you take the time to examine each piece, I'm confident you can see what each individual command is doing.

[Dimitris Kokkinos]

I could have never combine these functions by my own.I didn't even know about the function "reshape". The result is exactly what I was looking for.Thank you cyclist!

[Azzi Abdelmalek]

Dimitris, then accept his answer by clicking on [Accept this answer]

[James]

Hello,

I am facing a similar problem as the OP, however; I am only comparing each cell to a single value. Is there a faster way to do this? I have tried the suggested method and have found it to be nearly equivalent to a for loop method. Here is the example:

clear all
close all
clc

cID = false(100000,1); % generate logical vector
CELL = cell(100000, 1); % generating cell structure
for i = 1: length(CELL)-90000
    CELL{i} = randi(10, i, 1); % fill some cells with "random" integer
end

ID = 3; % index I'm looking for

% using for loop method ===================================================
tic
for i = 1: 10000 % looping over cells which contain values
    cID(i) = any(ismember(CELL{i}(:), ID)); % determine if cell contains ID
end
toc
% Elapsed time is 0.752470 seconds.

% alternatively using the cellfun method ==================================
tic
cID(1:10000) = cellfun(@(x) any(ismember(x,ID)),CELL(1:10000));
toc
% Elapsed time is 0.741742 seconds.

Any help on this topic would be greatly appreciated.

Thanks, James Peters

[Stephen23]

James: If the data arrays are contained in a cell array, then there you basically have two choices for comparing these with some scalar value:

• in a loop
• using cellfun

The loop may be a bit faster, but cellfun can produce much neater code.

Some other points in your code need improving though:

Note that because you are comparing two scalars there is no reason to use ismember at all: you can use logical relational operators for a scalar with any sized array:

idx = Array == Scalar;

If you are so concerned about speed why do you call randi 10,000 times in a loop? This indicates some reading up on MATLAB vectorization is required. For example, it could be generated faster and tidier using num2cell:

num2cell(randi(10,100,1))

This leads on to the most important question: why are you storing scalar values in a cell array? Unless you have a good reason to do this, you should keep your data in a numeric array, which then makes all the other operations suddenly much, much easier to perform... For example, all of your code could be reduced to:

A = randi(10,100,1);
B = A==3;

[James]

The only problem is that I have uneven arrays of different sizes. Hence the need for cells. However, using the logical == operator instead of the ismember function definitely gives better results. In fact it cut the time by about 3/4. I'm a little embarrassed I didn't think of it. Thank you for helping me out.

I'm actually writing my own Q-hull algorithm. This is for determining if a point is in the outside set of the current faces. I know MATLAB has its own Q-hull method built into the delaunayTriangulation function but I wanted to make my own.

Thanks, James Peters