search

GAMES-UChile / mogptk

Multi-Output Gaussian Process Toolkit
MIT License
161 stars 42 forks source link

Include input uncertainty in output estimation #67

Open manna88aero opened 11 months ago

manna88aero commented 11 months ago

At the moment the tool can estimate the uncertainty of the output given an input. But what happens when the input is also uncertain? In many practical cases the uncertainty in the input could be as large as the uncertainty of the output. Would it be possible to integrate input uncertainty to the tool?

felipe-tobar commented 10 months ago

Dear manna88aero,

Sorru for the delay. The models you refer to are called GP-LVM (Gaussian process latent variable model). In its current state, MOGPTK has not implemented GPLVMs, hopefully we can do it in the future.

Best, Felipe.

On 20-10-2023, at 06:02, manna88aero @.***> wrote:

At the moment the tool can estimate the uncertainty of the output given an input. But what happens when the input is also uncertain? In many practical cases the uncertainty in the input could be as large as the uncertainty of the output. Would it be possible to integrate input uncertainty to the tool?

— Reply to this email directly, view it on GitHub https://github.com/GAMES-UChile/mogptk/issues/67, or unsubscribe https://github.com/notifications/unsubscribe-auth/ACT3KG47RJPU7INDACAZUHTYAI42XAVCNFSM6AAAAAA6IWI7ECVHI2DSMVQWIX3LMV43ASLTON2WKOZRHE2TGOBVGIZDKNY. You are receiving this because you are subscribed to this thread.

tdewolff commented 9 months ago

On a side note, you can set Y_err when calling mogptk.Data(...) (see https://games-uchile.github.io/mogptk/data.html#mogptk.data.Data) to pass the standard deviation of the input data. When using the Exact model (the default), these will be passed as data_variance to https://games-uchile.github.io/mogptk/gpr/model.html#mogptk.gpr.model.Exact. Hopefully that satisfies your need.