Description Usage Arguments Value See Also Examples
Creates a kernel defining the properties of a gaussian process
1 2 3 | sde_kernel(type = c("exp_kernel", "rq_kernel", "sum_exp_kernels",
"exp_const_kernel", "clamped_exp_lin_kernel"), parameters,
inputDimension = 1, epsilon = 0)
|
type |
A string specifying the type of kernel to create |
parameters |
A NAMED list with the parameters that the kernel needs for its proper creation. The easiest way of seeing which parameters are required is to make this function fail. For example: sde_kernel("exp_kernel",list()) |
inputDimension |
The input dimension of the kernel |
epsilon |
A small value to be added to the diagonal of the kernel covariance matrices to regularize them. This improves the numerical stability of the computations. |
A sde_kernel S3 object
covmat
, autocovmat
, vars
,
get_hyperparams
, set_hyperparams
,
decrease_upper_bound
and increase_lower_bound
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 | # See demo/ornstein for a complete example
data("ornstein")
uncertainty = 5
inputDim = 1
# A small value to regularize covariance matrices
epsilon = 1e-5
# Create a exponential kernel with lengthScale = 1 and amplitude = uncertainty
# to model the drift function. We may select the amplitude from the fact that it
# is a gaussian process and the 95% confidence interval would be
# (-2 * sqrt(amplitude), 2 * sqrt(amplitude)).
driftKer = sde_kernel("exp_kernel",
list('amplitude' = uncertainty,
'lengthScales' = 1),
inputDim, epsilon)
# The selection of the amplitude parameters for gaussian process modelling the
# diffusion term is more complicated since voila actually uses a lognormal
# process. This function helps with the selection of the parameters from
# an uncertainty parameter and the time series
diffPars = select_diffusion_parameters(ornstein, deltat(ornstein),
priorOnSd = uncertainty)
# Create another exponential kernel with lengtScale=1
diffKer = sde_kernel("exp_kernel",
list('amplitude' = diffPars$kernelAmplitude,
'lengthScales' = 1),
inputDim, epsilon)
|
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.