Relationship between your LCM in Multi-output GP and LMC/ICM/SLFM

cornellius-gp / gpytorch

A highly efficient implementation of Gaussian Processes in PyTorch

MIT License

3.54k stars 557 forks source link

Relationship between your LCM in Multi-output GP and LMC/ICM/SLFM #700

Open ChunchaoPeter opened 5 years ago

ChunchaoPeter commented 5 years ago

Hi everyone,

I am a bit confused the LCM. However, I am familiar this paper: Kernels for Vector-Valued Functions: a Review.
https://arxiv.org/pdf/1106.6251.pdf

I can't quite tell whether the LCM kernel is the LMC kernel in the paper above. Could you please tell me what is the relationship between LCM and LMC?

Thanks in advance.

jacobrgardner commented 5 years ago

Yeah I think LCM is a typo for LMC; that's embarrassing :-)

I'll put up a PR that renames and soft deprecates LCM -> LMC.

ChunchaoPeter commented 5 years ago

Thanks for your response.

Balandat commented 5 years ago

I've seem LCM used as Linear Coregionalization Model rather than Linear Model of Coregionalization (LMC). It seems that the latter is more commonly used though.

ChunchaoPeter commented 5 years ago

Okok. Thanks for your response.

It seems Linear Coregionalization Model (LCM) in the paper Multitask Gaussian Processes Prediction: https://homepages.inf.ed.ac.uk/ckiw/postscript/multitaskGP_v22.pdf

I just wonder whether you plan to build a model for Linear Model of Coregionalization ( LMC) ?

Thanks.