search

carlosandrepaes / GaPP

Gaussian Processes GaPP code auxiliary repository
17 stars 11 forks source link

Problem when using observational data of the derivative of f(x) #2

Open vaneissyajp opened 3 years ago

vaneissyajp commented 3 years ago

I am using observational data of derivative of f(x) as prior. I have followed the example code that is given from the package's documentation. when I use it with my input data, there's bug. here is my code and my error. please let me know how to debug it. Thanks a lot in advance. image

carlosandrepaes commented 3 years ago

Good afternoon,

Thanks for your question. Indeed there is a bug in the gp and dgp modules that don't work in all numpy versions.

A quick fix that worked for me is the following:

Please change if (self.alpha == None): to if (self.alpha == None).all(): in gapp/gp.py and gapp/dgp.py

Hope it helps!

Best regards, Carlos

On Thu, 17 Jun 2021 at 11:24, vaneissyajp @.***> wrote:

I am using observational data of derivative of f(x) as prior. I have followed the example code that is given from the package's documentation. when I use it with my input data, there's bug. here is my code and my error [image: image] https://user-images.githubusercontent.com/81208271/122415728-5b6e0f80-cfb2-11eb-8a9d-d6d038ffc538.png

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/carlosandrepaes/GaPP/issues/2, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACMEIX3KDND5PRYEYQSCFN3TTIASHANCNFSM463ZHCNA .

vaneissyajp commented 3 years ago

Hello Carlos,

Thanks for the solution. I have changed the code as your recommendation, but I am still find the bug unsolved.

I have tried to change if (dX == None or dY == None or dSigma == None): to if (dX == None or dY == None or dSigma == None).any(): as the recommendation from jupyter but still get the error like this AttributeError: 'bool' object has no attribute 'any'

I found this bug when I add dX, dY, and dSigma in the parameter list. Please let me know how to debug it. Thanks in advance and have a nice day.

JCGoran commented 2 years ago

Hello, not-a-Gaussian-process-expert here, I randomly stumbled on this repo from the paper mentioned in the readme, since I also had trouble finding it at the original link (also, hi Carlos!). I think I fixed most of the problems mentioned here in my fork of this repo (based on igomezv/GaPP since it looked most recent), I'd be happy to open a PR to merge the changes if anyone would find it useful.