Error in NMSE = MSE/MSE00 matrix dimensions don't match

Hi all

In case the target and input have multiple rows , meaning , more than one input and target , calculating the MSE , I get one single number , and for MSE00 , I have a 3 *1 matrix , since I have 8 inputs and 3 Targets

how to solve it ?

Answer 1

[Greg Heath]

Apparently, you used the wrong equation. Both MSE and MSE00 are scalars.

Search

greg MSE00

[farzad]

Dear Professor

I have used

mean(var(t'),1)

and size(t') = 1055 3 .

that 1055 is the samples. so if this command is wrong , what should I use ?

[Greg Heath]

size(t) = [3 1055]

size(t') = [1055 3]

size(var(t')) = [ 1 3 ]

size(mean(var(t'))) = [ 1 1 ]

OK?

[farzad]

Yes Dear professor

but the above seems to be for MSE00a , not for MSE00 , right ?

[farzad]

I still find contradiction

in this link you have written MSE00 as :

MSE00=mean(var(t1,1)); % MSE00=0.0095

[farzad]

I used these lines and got the results that I have mentioned them besides the parameters, I would like to know if they are reasonable :

 Ntrneq = N*O    %3165
 ynaive = mean(t,2);    % 0.02*ones(O,1)
 Nw00   = numel(ynaive)       %  3
 Ndof00 = Ntrneq-Nw00         %  3162
 y00    = repmat(ynaive,1,N); % [3 1055]
 SSE00  = sse(t-y00)    %  1.2023e+04
 MSE00  = SSE00/Ntrneq        % 3.7988
 MSE00a  = SSE00/Ndof00       %  3.8024              
% MSE00  = mean(var(targets',1))  %  
% MSE00a = mean(var(targets',0))  % 
% LINEAR MODEL y0 = W0*[ones(1,N); inputs];

 W0    = t/[ones(1,N); x];
 Nw0   = numel(W0)               % 27
 Ndof0 = Ntrneq-Nw0              %  3138
 y0    = W0*[ones(1,N); x];
 SSE0  = sse(t-y0)         %  2.9892e+03 isn't this too big ???
 MSE0  = SSE0/Ntrneq             %   0.9445  biased
 MSE0a = SSE0/Ndof0              %   0.9526 DOFa                    
 R20   = 1-MSE0/MSE00            %   0.7514
 R20a  = 1-MSE0a/MSE00a          %   0.7495

the R2a in neural is 0.999 and mostly 1

I just wanted to know if these lines are correct

[farzad]

My data are 8 inputs and 3 outputs , I am wondering why my regression looks like this

[farzad]

I just don't know why the results are accumulated around the end part

[farzad]

It would be really good If I could know my numbers in the code make sense and are in an agreeable range and also if the ROC diagram is correct

[farzad]

could someone help me in evaluation of the posts above ?