Using fminsearch to determine variables

Currently I am working on estimating two different variables of a sigmoid curve graph, the first variable 'z(1)' is the slope of the sigmoid curve and 'z(2)' is half of the maximum height of the curve. I currently have an equation that outputs y-values of the curve, "E" based on these variables and the input/x value of 'C1': E=C1.^z(1)./(z(2)^z(1)+C1.^z(1)

I need to use fminsearch to calculate the values of 'z(1)' and 'z(2)'.

For the function tag I created:

f=@(z)C1.^z(1)./(z(2)^z(1)+C1.^z(1))

and then typed out

params=fminsearch(f,[1, 0.55])

with 1 and 0.55 being my guesses for the approximate half maximum value, and the slope of the sigmoid curve, respectively.

When I run the function it outputs an error on the fminsearch line I created: 'Unable to perform assignment because the size of the left side is 1-by-1 and the size of the right side is 41-by-1.'

Any help with fixing this would be appreciated!

Answer 1

[Star Strider]

The fminsearch function (and all other optimization functions I know of) want a scalar result from the function they are optimising.

In this instance, I would use:

f=@(z,C1) C1.^z(1)./(z(2)^z(1)+C1.^z(1));
params=fminsearch(@(z) norm(y-f(z,C1)),[1, 0.55])

That should work. If it has problems, please post the entire error message (if it throws an error), or a reasonably detailed description of what it is not doing that you want it to do, or what it is doing that you do not want it to do.

[Jordan Rosales]

Error message when using this code:

Unrecognized function or variable 'y'.

Error in rosalesLab7>@(z)norm(y-f(z,C1)) (line 100)

params=fminsearch(@(z) norm(y-f(z,C1)),[1, 0.55])

I am not sure how to define y so to get the function to run?

[Star Strider]

I assume here that ‘z’ is the parameter vector, ‘y’ is the dependent variable and that ‘C1’ is the independent variable vector.

If that is not the situation, please define them so I can make appropriate changes to my code example.

[Walter Roberson]

That is, in Star Strider's code, y stands for the vector of known output data values (must be a column vector) and C1 stands for the vector of known input data values (must be a column vector too)