Error:Assignment has more non-singleton rhs dimensions than non-singleton subscripts

So I have the following code to find the normal pdf of a given set of means(m) and standard deviations(s). Along with the 5th,25th,75th and 95th percentiles. x is a random variable with upper and lower limit.

 x=[2.22594:0.05:2.73436];

 m=[2.3890 2.3891 2.3782 2.4211 2.6127 2.7343 2.6273 2.4738 2.3638 2.2791 2.2259 2.2516];

 s=[0.0000 0.0000 0.0912 0.1416 0.1906 0.2200 0.2320 0.2378 0.2414 0.2453 0.2525 0.2791];

 y=zeros(1,12);

 z=zeros(1,12);

 for i=1:12

    y(1,i)=normpdf(x,m(1,i),s(1,i));

    plot(x,y);

    hold on

    z(1,i)=prctile(y,[5 25 75 95]);

 end

The error i am getting is "Assignment has more non-singleton rhs dimensions than non-singleton subscripts

Error in Math (line 8) y(1,i)=normpdf(x,m(1,i),s(1,i));" Can someone please help me solve this.

Answer 1

[Walter Roberson]

normpdf() generates one output value for each of its X input values. You then try to assign those multiple output values to a single location y(1,i) . You should probably use y(:,i) and the corresponding change to z, after changing the initialization of y to be zeros(length(x),length(m))

[Rishi Karia]

Ok so i did what you said and this is what showed up.

Edited code:

 x=[2.22594:0.05:2.73436];

 m=[2.3890 2.3891 2.3782 2.4211 2.6127 2.7343 2.6273 2.4738 2.3638 2.2791 2.2259 2.2516];
 s=[0.0000 0.0000 0.0912 0.1416 0.1906 0.2200 0.2320 0.2378 0.2414 0.2453 0.2525 0.2791];
 y=zeros(length(x),length(m));

 z=zeros(1,12);
 for i=1:12
    y(:,i)=normpdf(x,m(1,i),s(1,i));
    plot(x,y);
    hold on
    z(:,i)=prctile(y,[5 25 75 95]);
 end

Error:"Assignment has more non-singleton rhs dimensions than non-singleton subscripts

Error in Math (line 11) z(:,i)=prctile(y(:,i),[5 25 75 95]); "

[Rishi Karia]

OKay i changed the z variable in the loop to single variable and outputted it in every loop. It's fine now. Thanks a lot! @Walter Robinson.

[Nikhil Bharadwaj]

Hey! Even I have the same problem.I am working on Locally Weighted Linear Regression and I am facing the same problem .

for i = 1:m
  for j = 1:m
 W(j,j) = exp(-(X{1,i} - X{1,j}).^2 /(2 * tau * tau));

    end
    thetaweighted = pinv(Xfin' * W * Xfin) * Xfin' * W * Yfin;
    end

I am getting the error for W(j,j) = exp(-(X{1,i} - X{1,j}).^2 /(2 * tau * tau));

This code works for a single instance but not for a cell. Somebody help me ,I am new to matlab

[Walter Roberson]

What size do you expect the results to be? We can guess that your cell entries are not scalar.

[Nikhil Bharadwaj]

Theta weighted must be a 19*19 matrix.

[Walter Roberson]

I would prefer not to have to work through all the various matrix sizes that could lead to thetaweighted ending up being 19 x 19; I think there are multiple possibilities.

It would be much easier if you indicated what size you expect the X cell entries to be, and tau, and Xfin, and Yfin

[Nikhil Bharadwaj]

W is a 1*1 matrix and X is 18*163 , I want these dimensions to match so that i can find W and implement it in thetaweighted. Xfin and Yfin are already found .

[Walter Roberson]

I cannot help you until you tell me all of the array sizes involved, including the size of X{1,1} and tau. I also need to know what m is.

[M.Bilal]

Hey Walter, i have same problem and seeking your help in this regard.

I have a 6x6 matrix, say C_m, (as shown below) in which some entries are zero and all other entries are cloumn vectors (of order 142x1) c11, c12 & c44. eg c11 = [a11,a12,a13...], c12 = [b11,b12,b13...], c13 = [d11,d12,d13...], c44 =[e11,e12,13...].

How do I define all these column vectors in folllowing 6x6 matrix (C_m), so that matlab recognizes it?

And, all i need, is the resultant matrix C_m should be of the order 142x1 matrix.

C_m = [c11 c12 c12 0 0 0

c12 c11 c12 0 0 0

c12 c12 c11 0 0 0

0 0 0 2*c44 0 0

0 0 0 0 2*c44 0

0 0 0 0 0 2*c44]

Following problem occur, when i run the code? how do i solve this?

Error using horzcat
Dimensions of matrices being concatenated are not
consistent.

Error in bilal2 (line 287)
287      C_m =   [c11 c12 c12 0 0 0

[Walter Roberson]

Each of your hardcoded 0 constants need to be replaced by zeros() the same size as your c matrices.

    Z = zeros(size(c11));
    C_m =        [ c11 c12 c12 Z         Z     Z
                   c12 c11 c12 Z         Z     Z
                   c12 c12 c11 Z         Z     Z
                     Z   Z   Z 2*c44     Z     Z
                     Z   Z   Z Z     2*c44     Z
                     Z   Z   Z Z         Z 2*c44]

Answer 2

[Nikhil Bharadwaj]

I have attached the entire matlab file of my code.

[Walter Roberson]

X.mat and Y.mat ?

[Nikhil Bharadwaj]

my data

[Nikhil Bharadwaj]

These are the files.

[Walter Roberson]

Undefined function or variable 'compute_dtw_dist'.

The only reference I find to that routine is somewhere in the page https://wenku.baidu.com/view/193bddd5700abb68a982fb66.html which is a paper in Chinese whose name translates as "Research and application of time series anomaly detection"

[Nikhil Bharadwaj]

DTW file. The link u sent earlier is in chinese ..what do i do?

[Nikhil Bharadwaj]

I just want u to help me remove the error and run the file that's it .

[Walter Roberson]

Your X cell entries are arrays with 18 rows but a different number of columns from each other. What do you intend to be the result when you do

X{1,i} - X{1,j}

but X{1,i} and X{1,j} have a different number of columns?

You define W as a 10 x 10 numeric array, but you have

W{j,j} = exp(-(X{1,i} - X{1,j}).^2 /(2 * tau * tau))

which treats W as a cell array. The right hand side looks like it should be a numeric matrix (if only the sizes were the same...) rather than a scalar, so it does not appear that you could assign the result into a plain numeric array.

But later you have

    thetaweighted = pinv(Xfin' * W * Xfin) * Xfin' * W * Yfin;

which tries to use W as if it were a numeric array.

[Nikhil Bharadwaj]

The same code works good for a single instance and is not working good for a cell entry. Why is that?

[Nikhil Bharadwaj]

See this...this is the code for a single instance.I want the same results for LWRR.

[Nikhil Bharadwaj]

These are the data for single instance and the dimensions are 18*174 and 18*121 respectively.

[Walter Roberson]

"The same code works good for a single instance and is not working good for a cell entry. Why is that?"

Because the two sets of code do entirely different things. The version for one instance uses i and j as indices into a 2D array X, getting back a scalar for each indexing operation. The version with cells uses i and j as indices into a cell array, getting back an entire 2D array for each indexing operation.

At the moment your priority should be to get the cell version to work instead of trying to optimize it. So rewrite the cell version so it loops through the cell entries as a new level of for loop, extracting one particular cell entry into X, and doing the single-instance code with that X.

[Nikhil Bharadwaj]

Can you please help me rewrite the code?

[Walter Roberson]

I recommend that you learn to define functions. Make one from your LWR code, removing the part that loads the data, and instead assume that a single X and Y has been passed in, and have it compute and return whatever results you want. Then change your LWRR so that it loads the cell arrays from your files, and loops over them passing each individual entry pair to your function, and recording the result.

https://www.mathworks.com/help/matlab/ref/function.html

[Nikhil Bharadwaj]

Hey Walter , I need your help. I have 5 subjects each consisting of 460 sentences of 18*N dimension. 345 is for training and 115 is for testing,my job here is to convert the 18*Nn(Neutral) to 18*Nf(fast). I am applying Locally weighted linear regression to do the conversion in speech production . I need your advice on how I should proceed with my task.

[Walter Roberson]

I do not understand what you mean about that conversion ?

I am not familiar with locally weighted linear regression. I also do not know much about speech processing.

[Nikhil Bharadwaj]

No problem Walter you have been very helpful!