Mods to stochastic priors script

Author: mandar2812

scripts/stochasticPriors.scala

@@ -1,14 +1,15 @@
+import breeze.stats.distributions.Gaussian
 import io.github.mandar2812.dynaml.kernels.{MAKernel, RBFCovFunc}
 import io.github.mandar2812.dynaml.models.bayes.LinearTrendGaussianPrior
 import io.github.mandar2812.dynaml.probability.MultGaussianPRV
 import io.github.mandar2812.dynaml.analysis.implicits._
 import com.quantifind.charts.Highcharts._
 
 
-val rbfc = new RBFCovFunc(1.0)
-val n = new MAKernel(0.1)
+val rbfc = new RBFCovFunc(0.5)
+val n = new MAKernel(0.5)
 
-val gp_prior = new LinearTrendGaussianPrior[Double](rbfc, n, 0.5)
+val gp_prior = new LinearTrendGaussianPrior[Double](rbfc, n, 0.0)
 
 val xs = Seq.tabulate[Double](20)(0.5*_)
 
@@ -20,3 +21,17 @@ spline(xs, samples.head)
 hold()
 samples.tail.foreach((s: Seq[Double]) => spline(xs, s))
 unhold()
+
+val noiseAdd = new Gaussian(0.0, 0.1)
+
+val dataset = Seq((-0.1, 2.5 + noiseAdd.draw()), (10.51, 1.2 + noiseAdd.draw()))
+
+val gpModel = gp_prior.posteriorModel(dataset)
+val zs: MultGaussianPRV = gpModel.predictiveDistribution(xs)
+
+val samplesPost = zs.iid(8).sample().map(s => s.toBreezeVector.toArray.toSeq)
+
+spline(xs, samplesPost.head)
+hold()
+samplesPost.tail.foreach((s: Seq[Double]) => spline(xs, s))
+unhold()