Why do I get an error stating "Matrix dimensions must agree", whilst the dimensions appear to be equal to me.

Hi everyone,

I am runing the attached code, and for some reasons I get the error message stating:

Matrix dimensions must agree.

Error in ParametersFittingEulerMethodWater>EulerMethod (line 254)
  Results = sum((Results - xdata(i,:)).^2);

Error in fminsearch (line 200)
fv(:,1) = funfcn(x,varargin{:});

Error in ParametersFittingEulerMethodWater (line 58)
[b, Results] = fminsearch(@EulerMethod,b);

Whereas, to me, the dimensions appear to be equal. What am I doing wrong?

Please help.

[madhan ravi]

size(Results) %what does it show?
size(xdata)

[Dursman Mchabe]

Thanks a lot for the comment Madhan. Something is really strange. I get:

>

> size(xdata)
Undefined function or variable 'xdata'.
 
>> size(Results)
Undefined function or variable 'Results'.
 
>> 

[Dursman Mchabe]

Basically I need the sum of least squares on element by element, using:

for i = 1:length(tdata)
 %Results = sum((Results(i,:) - xdata(i,:)).^2);
 Results = sum(([Results(:,2) Results(:,4) Results(:,9)]) - ([xdata(:,2) xdata(:,3) xdata(:,4)])).^2;  
end

I cant figure how to do that.

[Image Analyst]

Dursman, you have to put the size() commands into your function, not in the command window (unless you're stopped at a breakpoint), because when the function exits, the variables are gone.

I also noted that it requires the Optimization Toolbox. For some reason, you didn't fill this out, or even the release you're using, when you posted, so I'll fill in the Toolbox for you so others don't waste time trying to run your code if they don't have that toolbox.

Answer 1

[Image Analyst]

When you do this:

Results - xdata(i,:)

then Results must be a column vector, not a row vector or 2-D matrix, because xdata(:, i) is a column vector. Basically : means "all", so xdata(:, i) means "All rows in column i".

Also we advise not to use i and j as variables since they also stand for the imaginary constant.

[Dursman Mchabe]

Ooops. Thanks a lot. I have just rectified that, the code is running well. It is just taking way too long. I hope it will give desired outcome.

[Stephen23]

"...the code is running well."

I doubt that, because how you defined the function EulerMethod does not make much sense. The fminsearch documentation describes the function as "fun is a function that accepts a vector or array x and returns a real scalar f (the objective function evaluated at x)", which means the x values are provided by fminsearch, and its scalar output depends on this. However while your function certainly has an input which you named b, you essentially ignore this input by entirely overwriting its values on the first lines of the function:

function Results = EulerMethod(b)
...
%% Parameters
b(1)        = 8.314;        
b(2) 		= 149;         
b(3)		= 5.15e+3;      
b(4) 		= 1.39e-9;      
b(5) 		= 2.89e-9;      
b(6) 		= 3.53e-9;      
b(7)		= 1.7e-3;       
b(8)		= 6.55e-8;      
b(9)		= 6.24;         
b(10)		= 5.68e-5;      
b(11)       = 5.3E-8 ;      
b(12)       = 3.1e-4;       
b(13)       = 10;           
b(14)		= 8.825e-3;     
b(15)       = 12.54;        
b(16)       = 100.0869;     
b(17)       = 8.4e-3;       
b(18)       = 2703;         
b(19)       = 258.30;       
b(20)       = 2540;         
b(21)       = 2e-08;        
b(22)       = 1.01E+05;     
b(23)       = 323.15;       
b(24)       = 4.e-4;        
b(25)       = 1.66667e-5;   
b(26)       = 2e-3;   
b(27)       = 3.257E-02;
b(28)       = 0.0;
b(29)       = 0.0;
b(30)       = 4.74e-5;      
b(31)       = 6e-4;       
b(32)       = 9.598e-5;

This means that regardless of the input values that fminsearch provides to your function it will ignore these values and immediately replace them with those hardcoded values. Which makes fminsearch totally useless.

Currently you define b, xdata, and tdata with exactly the same values in multiple locations, but you also declare them as global! Why did you declare pi as a global variable?

I recommend replacing those various global variables with a scalar structure:

https://www.mathworks.com/help/matlab/math/parameterizing-functions.html

[Dursman Mchabe]

Thanks a lot for the comment Stephen. The code had been running for hours now, yet the plots looks the same. fminsearch is indeed completely useless in this code. I will immediately try to pass b as a scalar structure and see what happens.

[Dursman Mchabe]

Hi Stephen. Please find the implementation of your recommendations on the newly-attached code. I hope I implemented every recommendation correctly.

(i) I have passed b as scalar structure.

(ii) I have commented-out b in the EulerMethod function, such that it does not use the hardcoded values.

(iii) I have also commented-out the repetitions and globals in multiple locations.

However the above changes lead to the error message:

Error using fminsearch (line 96)
FMINSEARCH accepts inputs only of data type double.
Error in ParametersFittingEulerMethodWater (line 59)
[b, Results] = fminsearch(@EulerMethod,b);

[Image Analyst]

So pass it what it wants - a double - not a structure.

[Dursman Mchabe]

Thanks for the comment Image Analyst. If I understand you correctly, you are implying that I try

[b, Results] = fminsearch(@EulerMethod,~);

I am understanding correctly?

[Stephen23]

"I have passed b as scalar structure."

Sorry, that was my mistake: I had one thing in my mind but wrote something else: it was those global variables that I would advise to convert to a scalar structure (or use nested functions). Then you can simplify/remove all of those global calls.

Of course the function input b must be a numeric array (e.g. a vector), because that is what fminserarch will provide to your function. So leave b exactly like you had it before.