Inner matrix dimensions must agree

I wrote these codes, I splitted the dataset(colon attached) into two parts (data1 and data2) based on the last column(1 or else)and make the columns equal to each others.

clc;
clear;
close all;
tic
load colon.mat 
data=colon; 
[n,m]=size(data);
d=10;
l=1;
t=1;
data1=[];
data2=[];
for i=1:n
    if data(i,m)==1
        data1(l,:)=data(i,1:m-1);
        l=l+1;
    else
        data2(t,:)=data(i,1:m-1); 
        t=t+1;
    end
end
if t>l
    data1(l:t-1,1:m-1)=0;
else
    data2(t:l-1,1:m-1)=0;
end
 for i=1: m-1                      
    thisCol1=data1(:,i);      
    thisCol2=data2(:,i);
    a24(i)=fkumarjohnson(thisCol1,thisCol2);
     end

and then in for loop I computed the distance value between the data1 and data2 and for computation I defined a function fkumarjohnson (I attached the image of the formula)but I found the error.

function [ fkj ] = fkumarjohnson(p,q )
%UNTITLED Summary of this function goes here
%   Detailed explanation goes here
fkj=sum(((p*p-q*q).^2)/(2*(dot(p,q)).^(3/2)));

end

the error is:

Error using  * 
Inner matrix dimensions must agree.

Error in fkumarjohnson (line 4)
fkj=sum(((p*p-q*q).^2)/(2*(dot(p,q)).^(3/2)));

Error in myFSmethod (line 62)
  a24(i)=fkumarjohnson(thisCol1,thisCol2);

The matlab said the inner matrix dimensions of two sides of * must agree but I don't know how to make their dimensions equal.I'll be very gratefull to have your opinions. Thanks

[Stephen23]

Are you sure that you want matrix multiplication?

Or do you actually need element-wise multiplication?

https://www.mathworks.com/help/matlab/matlab_prog/array-vs-matrix-operations.html

[phdcomputer Eng]

@Stephen Cobeldick Thanks, I want to calculate the result of the formula, and I guess that the formula needs element wise multipication because I'm computing the distance between every column of the matrix (colon) so all the elements in a column(all instances) is involved in the formula calculations. For element wise multiplication these codes are right?

Answer 1

[Fabio Freschi]

fkj= sum(.5*((p.^2-q.^2).^2)./((p.*q).^1.5));

Note that .^ .* and ./ operators are used for element-wise operation (see also Stephen's comment).

Note also that dot can give apparently unexpected results for complex inputs (see dot help)

[phdcomputer Eng]

@ Fabio Freschi Thanks, I used your advice and change the code of the formula :

fkj= sum(((p.^2-q.^2).^2)./(2.*((p.*q).^1.5)));

so you think than I should remove dot and replace with element wise operators?

[Fabio Freschi]

it is necessary to use element wise operators. you can leave the dot operator. But remember that for complex vectors, the dot product involves a complex conjugate and this could be not know to everyone.

[phdcomputer Eng]

Thanks greatly.

[phdcomputer Eng]

@Fabio Freschi Thanks greatly for your attention. the problem of the function in my question is solved by your advice, but for one function It has error:

function [ fkdg ] = fKDivergence( p,q )

fkdg=sum(p.*(log((2.*p)/(p+q))));

end

I used element-wise for several functions and their errors solved.

I used element-wise operator .* in the codes but MATLAB shows this error

Error using .*

Matrix dimensions must agree.

Error in fKDivergence (line 5)

fkdg=sum(p.*(log((2.*p)/(p+q))));

I'll be very gratefull to have your opinion. Thanks

Answer 2

[Image Analyst]

The error is described in the FAQ

[phdcomputer Eng]

@ Image AnalystThanks, but I didn't understand how to use cell array to solve the problem.

[Fabio Freschi]

Use ./ instead of /

[Image Analyst]

Cell arrays were the first FAQ entry. The link I gave you should have sent you directly to the error on "Inner dimensions must agree". It does for me. Perhaps you didn't give it enough time to scroll to the right part of the page.

Anyway the error was caused by you trying to do a matrix division instead of an element-by-element division. Using ./ like Fabio said will fix it.