| Author |
 Topic  |
|
|
AndreaP88
6 Posts |
 Posted - 10/17/2014 : 06:24:06 AM
|
Hi!
I would like to fit the following function
x=tanh(b*(J*x+h))
where b,J,h are parameters.
Is there a way to do this with origin?
Thanks in advance!
Origin Ver. and Service Release (Select Help-->About Origin): 9.1
Operating System: Windows 8.1 |
Edited by - AndreaP88 on 10/17/2014 06:24:49 AM |
|
|
jasonzhao
China
262 Posts |
Posted - 10/19/2014 : 11:27:11 PM
|
Dear Andrea,
Did you mean that y=tanh(b*(J*x+h)).
If it is, you can use user defined fitting function to solve this problem.
Please refer to the tutorial below for user defined fitting function.
http://www.originlab.com/doc/Tutorials/UserDef-FitFunc
and do not hesitate to contact us for further questions.
Best regards,
Jason Zhao
OriginLab Tech Service
|
 |
|
|
AndreaP88
6 Posts |
Posted - 10/21/2014 : 04:32:11 AM
|
Hi!
Thank you very much for your replay!
No actually the equation has x on both the right and left side! (this is what caused be some troubles)
All the best,
Andrea |
|
|
|
jasonzhao
China
262 Posts |
Posted - 10/21/2014 : 05:03:04 AM
|
Dear Andrea,
Regression attempts to model the relationship between two variables, However, in your model, there is only one variable X, so I do not really know which method should we adopt unless you tell me more details or thought about this unusual calculation.
Best regards,
Jason Zhao
OriginLab Tech Service |
Edited by - jasonzhao on 10/21/2014 05:05:16 AM |
|
|
|
Hideo Fujii
USA
1582 Posts |
Posted - 10/21/2014 : 3:29:52 PM
|
Hi Andrea,
Forgive me if I misunderstood what is your application as follows:
Your formula, x=tanh(b*(J*x+h)) leads the solution of simultaneous equation: y=x; and y=tanh(b*(J*x+h)) as shown in the sample below (here, b=3, J=2, h=-1 are given). I guess that when you have these solutions (in this sample case, (0.62155, 0.62155) and (0.9947, 0.9947), one may think that you can estimate the parameters from the solution.
Sure, you can apply fitting to this data with y=tanh(b*(J*x+h)) function. However, as depending on varying from a unique parameter set to infinitely many combinations, convergence may not be guaranteed.
For example, in this sample with given two datapoints, the function doesn't converge as overparameterized. In this sample, as shown, I tried to fix b and J to 2.5 and 1.5 respectively, and somehow converged to h=-0.64:
So, please consider carefully, if the above description makes some sense.
--Hideo Fujii
OriginLab |
|
|
| |
Topic |
|
|
|
Fitting self-consistency equation