(KRR) combines ridge regression (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. For non-linear kernels, this corresponds to a non-linear function in the original space. The form of the model learned by KRR is identical to support vector regression (SVR). However, different loss functions are used: KRR uses squared error loss while support vector regression uses epsilon-insensitive loss, both combined with l2 regularization. In contrast to SVR, fitting a KRR model can be done in closed-form and is typically faster for medium-sized datasets. On the other hand, the learned model is non-sparse and thus slower than SVR, which learns a sparse model for epsilon > 0, at prediction-time. This estimator has built-in support for multi-variate regression (i.e., when y is a 2d-array of shape [n_samples, n_targets]). Read more in the :ref:`User Guide <kernel_ridge>`.
ayemos
10tera
daichi_igarashi
jmatsu
fxwx23
thisaay
ktc_wada
inem
ippo012
oniak3ibm
yui_knk
kaspth
tkmnzm
philnash
destraynor
smashingmag
matthewcrist
malarkey
mojombo
holman
jrom
lara
samlambert
jeffersonlam
keavy
