Module Interfaces.Sigs.Eval.Covariances

module Covariances: sig .. end

Posterior covariances

type t

Type of covariances

val calc_model_inputs : Interfaces.Sigs.Eval.Model.t -> t

calc_model_inputs model
Returns covariances for all inputs used in model. This may be extremely expensive (O(N^2)) for large numbers of model inputs.

val calc : Interfaces.Sigs.Eval.Co_variance_predictor.t ->
       sigma2:float ->
       Interfaces.Sigs.Eval.Inputs.t -> t

calc co_variance_predictor ~sigma2 inputs
Returns posterior covariances for inputs given co_variance_predictor and noise level sigma2. This may be extremely expensive (O(N^2)) for large numbers of inputs.

val get : ?predictive:bool -> t -> Lacaml.D.mat

get ?predictive covariances
Returns the covariances as a matrix. If predictive is true, then the noise level will be added (to the diagonal only).

predictive : default = true

val get_variances : t -> Interfaces.Sigs.Eval.Variances.t

get_variances covariances
Returns the variances in covariances.