Hi pyGP Developers,
I am a newbie with GPs and I am trying to learn to use it for inference and predictions on real world datasets. My dataset is of the following format: [Latitude,Longitude,M1,...,M16] where M1...M16 represent the measurements recorded for 16 different years. The raw data has 59 locations so my data size is 59X18. I want to infer the signal_var,noise_var,latent length along latitude, latent length along longitude, latent length along time i.e. 5 parameters. I am not sure how to go about doing this mostly because my dataset seems to have heteroscedastic noise. If attaching the dataset would help , please let me know.
Any sort of help is appreciated.
hi @ktiwari9 , heteroscedastic gaussian process is unfortunately not implemented in pyGPs. You might want to take a look at GPy, where they have GPHeteroscedasticRegression.
Hi pyGP Developers, I am a newbie with GPs and I am trying to learn to use it for inference and predictions on real world datasets. My dataset is of the following format: [Latitude,Longitude,M1,...,M16] where M1...M16 represent the measurements recorded for 16 different years. The raw data has 59 locations so my data size is 59X18. I want to infer the signal_var,noise_var,latent length along latitude, latent length along longitude, latent length along time i.e. 5 parameters. I am not sure how to go about doing this mostly because my dataset seems to have heteroscedastic noise. If attaching the dataset would help , please let me know. Any sort of help is appreciated.