SumKernel

Defined in: ml/kernels/sum.js:11

Abstract base class for GP kernels

Extends

Kernel

Constructors

Constructor

new SumKernel(opts?): SumKernel

Defined in: ml/kernels/sum.js:16

Parameters

opts?

kernels

Kernel[]

Array of kernel instances to sum

Returns

SumKernel

Overrides

Kernel.constructor

Properties

kernels

kernels: Kernel[]

Defined in: ml/kernels/sum.js:27

Methods

call()

call(X1, X2?): Matrix

Defined in: ml/kernels/sum.js:42

Sum the children’s covariance MATRICES rather than their pointwise compute() values. Identical numbers for kernels that are plain functions of the input values, but a WhiteKernel is not one: it must know whether the matrix being built is K(X, X) or a cross-covariance K(X1, X2), which only call() can tell it. Delegating per element would silently drop the noise term (or, worse, leak it into the train/test block).