GoGP is a library for probabilistic programming around Gaussian processes. It uses Infergo for automatic differentiation and inference.

GoGP is built around a dual view on the Gaussian process

• as a stochastic process,
• as a probabilistic model with respect to kernel.

Gaussian process instance

GP, the Gaussian process type, encapsulates similarity and noise kernels, their parameters, and observation inputs and outputs:

// Type GP is the barebone implementation of GP.
type GP struct {
	NDim                   int       // number of dimensions
	Simil, Noise           Kernel    // kernels
	ThetaSimil, ThetaNoise []float64 // kernel parameters

	X [][]float64 // inputs
	Y []float64   // outputs

	Parallel bool // when true, covariances are computed in parallel
}

Public methods defined on GP fall into two groups: Gaussian process fitting and prediction, on one hand, and probabilistic model interface, on the other hand.