search

EmuKit / emukit

A Python-based toolbox of various methods in decision making, uncertainty quantification and statistical emulation: multi-fidelity, experimental design, Bayesian optimisation, Bayesian quadrature, etc.
https://emukit.github.io/emukit/
Apache License 2.0
605 stars 128 forks source link

Constraining multi-fidelity gp model #438

Closed kenzaxtazi closed 6 months ago

kenzaxtazi commented 1 year ago

Hi,

I'm looking to apply the multi-fidelity GP model to downscaling climate variables and was wondering the best way to constrain or set priors on the hyperparameters.

This is an example of the output I currently have. The stripes along the same longitudes ('lon') are not physical and I think the model should be nudged in the right direction. I'm using the linear model here with two fidelities and 4 dimensions (time, lat, lon, elevation) with ARD.

I've tried constraining and setting priors on the hyperparameters and get more realistic outputs but significantly different estimates for the values across my study area.

Screenshot 2023-02-26 at 15 45 08
apaleyes commented 1 year ago

Hi @kenzaxtazi ! Not sure if I fully understand the question. Can you maybe post what you've tried and what's not working? Also, what are your fidelities?

apaleyes commented 6 months ago

Closing due to inactivity, feel free to reopen or ask a new one if you'd like to continue the conversation!