File Exchange

image thumbnail

sigm_fit

version 1.5.0.0 (4.11 KB) by R P
Optimization of parameters of the sigmoid function

77 Downloads

Updated 30 Mar 2016

View License

Optimization of parameters of the sigmoid function

Syntax:
[param]=sigm_fit(x,y)

that is the same that
[param]=sigm_fit(x,y,[],[],[]) % no fixed_params, automatic initial_params

[param]=sigm_fit(x,y,fixed_params) % automatic initial_params
[param]=sigm_fit(x,y,[],initial_params) % use it when the estimation is poor
[param]=sigm_fit(x,y,fixed_params,initial_params,plot_flag)

param = [min, max, x50, slope]

if fixed_params=[NaN, NaN , NaN , NaN] % or fixed_params=[]
optimization of "min", "max", "x50" and "slope" (default)

if fixed_params=[0, 1 , NaN , NaN]
optimization of x50 and slope of a sigmoid of ranging from 0 to 1


Additional information in the second output, STAT
[param,stat]=sigm_fit(x,y,fixed_params,initial_params,plot_flag)


Example:
%% generate data vectors (x and y)
fsigm = @(param,xval) param(1)+(param(2)-param(1))./(1+10.^((param(3)-xval)*param(4)))
param=[0 1 5 1]; % "min", "max", "x50", "slope"
x=0:0.1:10;
y=fsigm(param,x) + 0.1*randn(size(x));

%% standard parameter estimation
[estimated_params]=sigm_fit(x,y)

%% parameter estimation with forced 0.5 fixed min
[estimated_params]=sigm_fit(x,y,[0.5 NaN NaN NaN])

%% parameter estimation without plotting
[estimated_params]=sigm_fit(x,y,[],[],0)

Comments and Ratings (21)

Oded Scharf

Swati Jain

Great function!

Bram Mast

GDMFSOB Bloody brilliant! Thank you.

Sam Garrard

Zoe

RJ Ellis

Works right out of the box. Thanks!

Reza

Works as expected!

Simple and effective. Similarly to polyfit-polyval, it would be great to include the related sigm_val function, to evaluate the generated sigmoidal curve.

noam josef

tried using the function with my data and the givven example but in both recieved "Undefined function 'sigm_fit' for input arguments of type
'double'."

any thughts about it?

Yongshen

This requires a minmax() function - which toolbox is that in?

Updates

1.5.0.0

title update

1.5.0.0

plot bug fixed

1.4.0.0

minmax function used in plotting substituted by min and max functions

1.3.0.0

STAT output variable, with confidence intervals for parameters and distribution

MATLAB Release Compatibility
Created with R13
Compatible with any release
Platform Compatibility
Windows macOS Linux
Acknowledgements

Inspired by: Sine function fit

Inspired: sigmoid, sigm_fit_val(param,x)