Hello I wanted to write code in matlab "without using any toolbox" to do the following things, can we do like this?

This is the equation calc_value = C0+C1*N+C2*N^2+C3*P1+C4*P2

where C0-C4 = coefficients N, P1, P2 = input values Now suppose the actual value = x

So, error = calc_value - actual value(x)

I squared the error = error^2

Now I want to minimize the sum of error^2 by adjusting the coefficients C0-C4 in the above equations (like we use solver in excel)

Note: Here the actual values and the inputs are available, I don't want to use any toolboxes

this is my table

N N^2 P1 P2 calc actual error error^2

please help if anyone has the idea...

Thanks Vishal

[dpb]

You're trying to find five constants with only one equation -- you need at least that many independent measurements.

[Image Analyst]

How can you have an error (meaning calc-actual) if you don't have the coefficients yet?

[Vishal Halale]

At start I am assuming the coefficients c0-c4 = 1 then with the help of that I am calculating the value and with using those values finding the error. Then again minimizing the error by adjusting the assumed coefficients. Did you get it?

Answer 1

[Image Analyst]

Can't you just use least squares, like

coefficients = calc_value / yourTable;

or something like that? Of course the table would not include calc, actual, error, and must have a column of 1's to start

1, N1, N1^2, P11, P21;
1, N2, N2^2, P12, P22;
1, N3, N3^2, P13, P23;
etc.

Or use the standard least squares solution

Where X is the table I gave above, and y is your calc_values, and beta is your Cn coefficients.

[Vishal Halale]

Hey Thanks for the reply. Here See this Image It will give you the clear idea what I wanted. Please write some code also if you know, coz I am new to matlab. All I wanted out of this table is the correct/matlab calculated values of c0-c4 which will minimize the error^2.

Thanks.