Combine kernels hierarchy
neps.optimizers.bayesian_optimization.kernels.combine_kernels_hierarchy
#
SumKernel
#
Bases: CombineKernel
Source code in neps/optimizers/bayesian_optimization/kernels/combine_kernels_hierarchy.py
forward_t
#
forward_t(
weights: Tensor,
gr2: list,
x2=None,
gr1: list = None,
x1=None,
feature_lengthscale=None,
)
Compute the kernel gradient w.r.t the feature vector Parameters
feature_lengthscale x2 x1 gr1 weights gr2
Returns#
grads: k list of 2-tuple. (K, x2) where K is the weighted Gram matrix of that matrix, x2 is the leaf variable on which Jacobian-vector product to be computed.