[Tutor] Gaussian process regression

[Steven D'Aprano]

On Fri, Feb 27, 2015 at 03:04:25AM +0900, Huijae wrote:

> Hi, I am trying to use Gaussian process regression for Near Infrared 
> spectra. I have reference data(spectra), concentrations of reference 
> data and sample data, and I am trying to predict concentrations of 
> sample data. Here is my code. from sklearn.gaussian_process import 
> GaussianProcess gp = GaussianProcess() gp.fit(reference, 
> concentration) concentration_pred = gp.predict(sample) The results 
> always gave me the same concentration even though I used different 
> sample data. When I used some parts of reference data as sample data, 
> it predicted concentration well. But whenever I use different data 
> than reference data, it always gave me the same concentration. Can I 
> get some help with this problem? What am I doing wrong? I would 
> appreciate any help. Thanks, Jay

Your mail program has mangled your code and mixed up the layout. I 
suggest turning off all formatting, "Rich Text" or HTML mail, since that 
usually destroys the layout of code.

I'm going to try to guess your code's layout:

from sklearn.gaussian_process import GaussianProcess 
gp = GaussianProcess() 
gp.fit(reference, concentration) 
concentration_pred = gp.predict(sample)


This is not enough for us to give any meaningful advice. What is 
"sklearn"? What are "reference", "concentration", "sample"? We'd need 
to see actual data, and know what GaussianProcess does, to help.


> The results always gave me the same concentration 
> even though I used different sample data.

*Exactly* the same? To 17 decimal places? Or just close to the same?

Perhaps the different samples give the same result because they are 
samples from the same population.


> When I used some parts of 
> reference data as sample data, it predicted concentration well. But 
> whenever I use different data than reference data, it always gave me 
> the same concentration.

You may need to ask people who know more about GaussianProcess and how 
it works.



-- 
Steve