Unable to perform assignment because the indices on the left side are not compatible with the size of the right side

Hello, I am getting this error whenever I try to run this.

Problem: NN1 is my data set, column 1-5 are inputs and 7th is the target. Funny thing is that it works perfectly well when there were half of the data, but even if i add 1 more row of data in the input and target column, this error is displayed.

y and y2 infos:

y is 132x1 double

y2 is 132x1 complex double

Thanks in advance

data = readmatrix('NN1.xlsx');
x = data(:,1:5);
y = data(:,7);
m = length(y);
histogram(x(:,2),50);
plot(x(:,1),y,'o')

y2 = log(1+y);
for i = 1:5
    x2(:,i) = (x(:,i)-min(x(:,i)))/(max(x(:,i))-min(x(:,i)));      %Error here 
end
histogram(x2(:,5),50);
plot(x2(:,5),y2,'o');

[Walter Roberson]

When you add the one more row... do you clear x2 ?

[Mehbubul Haque]

Hi, thank you, I did not before. But later it worked.

Answer 1

[Jon]

Following up on @Walter Roberson's comment, I ran your code (I assigned my own values to data, as I didn't have your input file) and found that it can run repeatedly as long as you don't change the number of rows in data. If you run the code first say with data having 132 rows, and then run it again, without clearing the workspace with data having 133 rows you get that error.

This makes sense as the first time through the loop on the second run x2 is already a 132 x 5 matrix. But you are trying to assign a 133 x 1 vector to its first column. This won't work.

You could simply clear the workspace, or just x2 between runs, but the real problem here is that you are not preallocating your arrays. If you do this as shown below it will use MATLAB's memory more efficiently, and also avoid these type of problems with assigning into arrays that already have values from some previous run. Please also seehttps://www.mathworks.com/help/matlab/matlab_prog/preallocating-arrays.html

data = rand(132,7)
x = data(:,1:5);
y = data(:,7);
m = length(y);
histogram(x(:,2),50);
plot(x(:,1),y,'o')

y2 = log(1+y);
% preallocate array to hold results
numRows = size(data,1);
x2 = zeros(numRows,5);
for i = 1:5
    x2(:,i) = (x(:,i)-min(x(:,i)))/(max(x(:,i))-min(x(:,i)));      %Error here 
end
histogram(x2(:,5),50);
plot(x2(:,5),y2,'o');

Finally, in your particular example, you don't need to use a loop or preallocate at all. You can do everything in a "vectorized" fashion using just the one line of code:

x2 = (x - min(x))./(max(x) - min(x)) % note the ./ to give element by element division

[Walter Roberson]

x2 = rescale(x) ;

[Jon]

Not quite clear what you are suggesting. Does the scaling itself have any benefits or would it work just as well to initialize?

x2 = zeros(size(x))

Is the scaling irrelevant, but just a trick to do it with just the simple x as the argument rather than size(x) as above?

Anyhow I guess real point is that you should always initialize the matrix before assigning submatrices of it in a loop

[Walter Roberson]

x = randi([100, 2000], 1, 20)

x = 1×20

325 174 998 1742 1115 267 963 1656 283 1411 1094 143 1632 1457 292 1314 531 1260 181 1531

x2 = (x - min(x))./(max(x) - min(x))

x2 = 1×20

0.1138 0.0194 0.5347 1.0000 0.6079 0.0775 0.5128 0.9462 0.0876 0.7930 0.5947 0 0.9312 0.8218 0.0932 0.7323 0.2427 0.6986 0.0238 0.8680

x3 = rescale(x)

x3 = 1×20

0.1138 0.0194 0.5347 1.0000 0.6079 0.0775 0.5128 0.9462 0.0876 0.7930 0.5947 0 0.9312 0.8218 0.0932 0.7323 0.2427 0.6986 0.0238 0.8680

Notice that x2 and x3 are the same. That is, instead of using a loop, or using the vectorized form that you correctly suggested, you can just call rescale() to do this particular mathematical operation (and do not need to pre-allocate either if you call it.)

[Jon]

Oh, I missed that it actually did the operation the OP was asking for. That's really nice! Thanks that's a nice useful function to know about.

[Mehbubul Haque]

Thank you for the answers, I got it now.