Samplers

struct Sampler{RNG<:Random.AbstractRNG}

Wraps the parameter distributions and random number generator used to draw random feature parameters.

• parameter_distribution::EnsembleKalmanProcesses.ParameterDistributions.ParameterDistribution: A probability distribution, possibly with constraints

• rng::Random.AbstractRNG: A random number generator state

Constructors

• FeatureSampler(parameter_distribution, bias_distribution; rng) — explicit ParameterDistribution bias, or nothing for no bias term.
• FeatureSampler(parameter_distribution, output_dim; uniform_shift_bounds, rng) — uniform bias over output_dim dimensions; bounds default to [0, 2π].

FeatureSampler(
    parameter_distribution::EnsembleKalmanProcesses.ParameterDistributions.ParameterDistribution,
    bias_distribution::Union{Nothing, EnsembleKalmanProcesses.ParameterDistributions.ParameterDistribution};
    rng
) -> Sampler{Random.TaskLocalRNG}

Construct a Sampler from a parameter_distribution and an optional bias_distribution.

When bias_distribution is nothing, no bias term is added to the feature inner product. When a ParameterDistribution is supplied, it is combined with parameter_distribution into a single joint distribution before wrapping in the Sampler.

Pass a ParameterDistribution with name "xi" for the feature weights.

get_parameter_distribution(
    s::Sampler
) -> EnsembleKalmanProcesses.ParameterDistributions.ParameterDistribution

gets the parameter_distribution field

get_rng(s::Sampler) -> Random.AbstractRNG

gets the rng field

sample(
    rng::Random.AbstractRNG,
    s::Sampler,
    n_draws::Int64
) -> EnsembleKalmanProcesses.ParameterDistributions.ParameterDistribution

samples the distribution within s, n_draws times using a random number generator rng. Can be called without rng (defaults to s.rng) or n_draws (defaults to 1)