Model and Trace

Model and Trace are both found in com.stripe.rainier.core.

Model

Instance Methods

  • prior: Model

Strips away the observations, for ease of checking prior predictions.

  • merge(other: Model): Model

Combines two models.

  • sample(config: SamplerConfig, nChains: Int = 4): Trace

Run inference using the provided sampler configuration.

  • optimize[T](value: Generator[T]): T

Run L-BFGS. Note that this method will accept non-Generator values, and automatically wrap them with Generator(), if possible.

Object Methods

  • observe[Y](ys: Seq[Y], likelihood: Distribution[Y]): Model
  • observe[Y](ys: Seq[Y], likelihoods: Vec[Distribution[Y]): Model

Trace

  • diagnostics: List[Trace.Diagnostics]

Produce a list of Diagnostics(rHat: Double, effectiveSampleSize: Double), one for each parameter. Requires chains > 1.

  • thin(n: Int): Trace

Keep every n'th sample in each chain.

  • predict[T](value: Generator[T]): List[T]

Generate one value from generator from each sample in the trace. Like optimize, this will automatically convert values into Generator where possible.

  • mean[N:Numeric](value: Generator[N]): Double

For any numeric Generator, compute the posterior expectation.

← DistributionsGenerator →