Copyright	(c) Adam Scibior 2015-2020
License	MIT
Maintainer	leonhard.markert@tweag.io
Stability	experimental
Portability	GHC
Safe Haskell	Safe-Inferred
Language	Haskell2010

Control.Monad.Bayes.Traced.Static

Description

Synopsis

data Traced m a = Traced {
- model :: Weighted (Density m) a
- traceDist :: m (Trace a)
}
hoist :: (forall x. m x -> m x) -> Traced m a -> Traced m a
marginal :: Monad m => Traced m a -> m a
mhStep :: MonadDistribution m => Traced m a -> Traced m a
mh :: MonadDistribution m => Int -> Traced m a -> m [a]

Documentation

data Traced m a Source #

A tracing monad where only a subset of random choices are traced.

The random choices that are not to be traced should be lifted from the transformed monad.

Constructors

Traced
Fields model :: Weighted (Density m) a traceDist :: m (Trace a)

Instances

Instances details

MonadTrans Traced Source #
Instance details Defined in Control.Monad.Bayes.Traced.Static Methods lift :: Monad m => m a -> Traced m a #
Monad m => Applicative (Traced m) Source #
Instance details Defined in Control.Monad.Bayes.Traced.Static Methods pure :: a -> Traced m a # (<>) :: Traced m (a -> b) -> Traced m a -> Traced m b # liftA2 :: (a -> b -> c) -> Traced m a -> Traced m b -> Traced m c # (>) :: Traced m a -> Traced m b -> Traced m b # (<*) :: Traced m a -> Traced m b -> Traced m a #
Monad m => Functor (Traced m) Source #
Instance details Defined in Control.Monad.Bayes.Traced.Static Methods fmap :: (a -> b) -> Traced m a -> Traced m b # (<$) :: a -> Traced m b -> Traced m a #
Monad m => Monad (Traced m) Source #
Instance details Defined in Control.Monad.Bayes.Traced.Static Methods (>>=) :: Traced m a -> (a -> Traced m b) -> Traced m b # (>>) :: Traced m a -> Traced m b -> Traced m b # return :: a -> Traced m a #
MonadDistribution m => MonadDistribution (Traced m) Source #
Instance details Defined in Control.Monad.Bayes.Traced.Static Methods random :: Traced m Double Source # uniform :: Double -> Double -> Traced m Double Source # normal :: Double -> Double -> Traced m Double Source # gamma :: Double -> Double -> Traced m Double Source # beta :: Double -> Double -> Traced m Double Source # bernoulli :: Double -> Traced m Bool Source # categorical :: Vector v Double => v Double -> Traced m Int Source # logCategorical :: (Vector v (Log Double), Vector v Double) => v (Log Double) -> Traced m Int Source # uniformD :: [a] -> Traced m a Source # geometric :: Double -> Traced m Int Source # poisson :: Double -> Traced m Int Source # dirichlet :: Vector v Double => v Double -> Traced m (v Double) Source #
MonadFactor m => MonadFactor (Traced m) Source #
Instance details Defined in Control.Monad.Bayes.Traced.Static Methods score :: Log Double -> Traced m () Source #
MonadMeasure m => MonadMeasure (Traced m) Source #
Instance details Defined in Control.Monad.Bayes.Traced.Static

hoist :: (forall x. m x -> m x) -> Traced m a -> Traced m a Source #

marginal :: Monad m => Traced m a -> m a Source #

Discard the trace and supporting infrastructure.

mhStep :: MonadDistribution m => Traced m a -> Traced m a Source #

A single step of the Trace Metropolis-Hastings algorithm.

mh :: MonadDistribution m => Int -> Traced m a -> m [a] Source #

Full run of the Trace Metropolis-Hastings algorithm with a specified number of steps. Newest samples are at the head of the list.

For example:

I have forgotten what day it is.
There are ten buses per hour in the week and three buses per hour at the weekend.
I observe four buses in a given hour.
What is the probability that it is the weekend?

>>> :{
 let
   bus = do x <- bernoulli (2/7)
            let rate = if x then 3 else 10
            factor $ poissonPdf rate 4
            return x
   mhRunBusSingleObs = do
     let nSamples = 2
     sampleIOfixed $ unweighted $ mh nSamples bus
 in mhRunBusSingleObs
:}
[True,True,True]

Of course, it will need to be run more than twice to get a reasonable estimate.