Methods to construct, combine and evaluate kernels (covariance functions)
# static kernel objects
k <- kernels$White(input_dim, variance = 1, active_dims = NULL)
k <- kernels$Constant(input_dim, variance = 1, active_dims = NULL)
k <- kernels$Bias(input_dim, variance = 1, active_dims = NULL)
# stationary kernel objects
k <- kernels$RBF(input_dim, variance = 1, lengthscales = NULL, active_dims = NULL, ARD = FALSE)
k <- kernels$Exponential(input_dim, variance = 1, lengthscales = NULL, active_dims = NULL, ARD = FALSE)
k <- kernels$Matern12(input_dim, variance = 1, lengthscales = NULL, active_dims = NULL, ARD = FALSE)
k <- kernels$Matern32(input_dim, variance = 1, lengthscales = NULL, active_dims = NULL, ARD = FALSE)
k <- kernels$Matern52(input_dim, variance = 1, lengthscales = NULL, active_dims = NULL, ARD = FALSE)
# non-stationary kernel objects
k <- kernels$Linear(input_dim, variance = 1, lengthscales = NULL, active_dims = NULL, ARD = FALSE)
# periodic kernel objects
k <- kernels$Cosine(input_dim, variance = 1, lengthscales = NULL, active_dims = NULL, ARD = FALSE)
k <- kernels$PeriodicKernel(input_dim, period = 1, variance = 1, lengthscales = NULL, active_dims = NULL)
# kernel operations
k + k
k * k
# kernel object member functions
k$K(X, X2 = NULL)
k$Kdiag(X)
k$compute_K(X, Z)
k$compute_K_symm(X)
#k$compute_Kdiag(X)
An integer vector giving the dimensions of the matrix
X on which member functions will operate.
An
integer vector identifying the columns of X on which this kernel
operates. If NULL (default), all columns are active
A positive numeric scalar giving the initial value of the marginal variance of the kernel.
A positive numeric vector giving the initial
value of kernel lengthscales for the columns indexed by active_dims.
A positive numeric vector giving the initial value of the
periodicity for the columns indexed by active_dims.
Whether allow lengthscales and periodicities to vary between
active dimensions. If FALSE, lengthscales and period
should be scalar.