Menu
I'm using the KernelRidge estimator, which apparently supports various kernels. I've read the documentation:
Apparently one can define own kernel functions. Somehow I have the impression, that there are some predefined kernels. Unfortunately, I can't them? Do you know if there are some? And if yes, could you point me to a location which describes them and their parameters?
999
Kernel ridge regression (KRR) [M2012] combines Ridge regression and classification (linear least squares with l2-norm regularization) with the kernel trick. It thus learns a linear function in the space induced by the respective kernel and the data.
Kernel ridge regression models are nonparametric regression models that are capable of modeling linear and nonlinear relationships between predictor variables and outcomes. Results can be highly sensitive to choices of model hyperparameters.
One of the few examples of metric learning algorithms focused specifically on regres- sion tasks is Metric Learning for Kernel Regression (MLKR) developed by Weinberger and Tesauro [45].
Following kernels are supported: RBF, laplacian, polynomial, exponential, chi2 and sigmoid kernels. I drew conclusion from observing the "gamma parameter" description of KernelRidge documentation.
The mathematical formulation of these kernels can be found at this link as mentioned earlier by @ndrizza.
I am also trying to figure out the string arguments for kernel, based on information from SVR page, I think they are something like linear, poly, rbf, and sigmoid. SVR page.
183
I found a good resource.
Here is a list of some kernels:
http://scikit-learn.org/stable/modules/classes.html#module-sklearn.metrics.pairwise
Here are some further details about the kernels:
http://scikit-learn.org/stable/modules/metrics.html#metrics
34
If you love us? You can donate to us via Paypal or buy me a coffee so we can maintain and grow! Thank you!
Donate Us With
