Timeline for Gaussian Process Kernel and Ridge Regression
Current License: CC BY-SA 3.0
9 events
| when toggle format | what | by | license | comment | |
|---|---|---|---|---|---|
| Apr 5, 2014 at 16:32 | comment | added | usεr11852 | Good answer (+1) but can you comment a bit further on how coding affects the relation? Why isn't the $K = K+\lambda I$-like the "universal relation behind this regularization idea? Unless we are talking about a Generalized Tikhonov regularization scheme (which OK, it is a superset of ridge regression) I can't see why coding would be an issue. (If this happens to be a long answer feel free to point me to papers) | |
| Mar 20, 2014 at 18:58 | vote | accept | papajohn | ||
| Mar 19, 2014 at 17:01 | comment | added | Dikran Marsupial | I use leave-one-out cross-validation for setting the regularisation parameter of ridge regression models, which can be performed very cheaply (see dx.doi.org/10.1016/j.neunet.2007.05.005 ). However, IIRC, the regularisation parameter for ridge regression typically corresponds to a parameter of the covariance function of the GP, however the exact relationship depends on how it is coded. | |
| Mar 19, 2014 at 16:56 | comment | added | papajohn | CV Makes sense... but my objective is to simulate the GP so how would the choice be done in that case? | |
| Mar 19, 2014 at 16:51 | comment | added | Marc Claesen | @papajohn typically through cross-validation. | |
| Mar 19, 2014 at 15:17 | comment | added | papajohn | thanks I ll give it a try with theano and python! How would somebody chose the regularisation parameter? Bayesian optimisation until it fits? | |
| Mar 19, 2014 at 15:07 | comment | added | Dikran Marsupial | use the same kernel as the covariance function of the GP, there is also a need to choose a suitable regularisation parameter for the KRR model to match the GP. | |
| Mar 19, 2014 at 14:11 | history | edited | Dikran Marsupial | CC BY-SA 3.0 |
added 200 characters in body
|
| Mar 19, 2014 at 14:06 | history | answered | Dikran Marsupial | CC BY-SA 3.0 |