Added types to methods of ModelTuner hierarchy

 - Configuration methods now return `this.type`

Author: mandar2812

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/optimization/AbstractCSA.scala

@@ -26,21 +26,21 @@ import scala.util.Random
 
 /**
   * Skeleton implementation of the Coupled Simulated Annealing algorithm
-  * @author mandar datum 01/12/15.
-  */
+  * @author mandar2812 date 01/12/15.
+  * */
 abstract class AbstractCSA[M <: GloballyOptimizable, M1](model: M)
   extends AbstractGridSearch[M, M1](model: M) {
 
   protected var MAX_ITERATIONS: Int = 10
 
   protected var variant = AbstractCSA.MuSA
 
-  def setVariant(v: String) = {
+  def setVariant(v: String): this.type = {
     variant = v
     this
   }
 
-  def setMaxIterations(m: Int) = {
+  def setMaxIterations(m: Int): this.type = {
     MAX_ITERATIONS = m
     this
   }
@@ -116,7 +116,7 @@ abstract class AbstractCSA[M <: GloballyOptimizable, M1](model: M)
     def CSATRec(eLandscape: List[(Double, Map[String, Double])], it: Int): List[(Double, Map[String, Double])] =
       it match {
         case 0 => eLandscape
-        case num =>
+        case _ =>
           logger.info("**************************")
           logger.info("CSA Iteration: "+(MAX_ITERATIONS-it+1))
           //mutate each element of the grid with

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/optimization/MixtureMachine.scala

@@ -65,11 +65,14 @@ BaseProcess <: ContinuousProcessModel[T, I, Y, W1]
     this
   }
 
-  protected def calculateEnergyLandscape(initialConfig: Map[String, Double], options: Map[String, String]) =
+  protected def calculateEnergyLandscape(
+    initialConfig: Map[String, Double],
+    options: Map[String, String]): Seq[(Double, Map[String, Double])] =
     if(policy == "CSA") performCSA(initialConfig, options)
     else getEnergyLandscape(initialConfig, options, meanFieldPrior)
 
-  protected def modelProbabilities = DataPipe(ProbGPCommMachine.calculateModelWeightsSigmoid(baselinePolicy) _)
+  protected def modelProbabilities: DataPipe[Seq[(Double, Map[String, Double])], Seq[(Double, Map[String, Double])]] =
+    DataPipe(ProbGPCommMachine.calculateModelWeightsSigmoid(baselinePolicy))
 
   override def optimize(
     initialConfig: Map[String, Double],

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/optimization/ModelTuner.scala

@@ -44,27 +44,27 @@ trait ModelTuner[T <: GloballyOptimizable, T1] {
 
   protected var meanFieldPrior: Map[String, ContinuousRVWithDistr[Double, ContinuousDistr[Double]]] = Map()
 
-  def setPrior(p: Map[String, ContinuousRVWithDistr[Double, ContinuousDistr[Double]]]) = {
+  def setPrior(p: Map[String, ContinuousRVWithDistr[Double, ContinuousDistr[Double]]]): this.type = {
     meanFieldPrior = p
     this
   }
 
-  def setNumSamples(n: Int) = {
+  def setNumSamples(n: Int): this.type = {
     num_samples = n
     this
   }
 
-  def setLogScale(t: Boolean) = {
+  def setLogScale(t: Boolean): this.type = {
     logarithmicScale = t
     this
   }
 
-  def setGridSize(s: Int) = {
+  def setGridSize(s: Int): this.type = {
     this.gridsize = s
     this
   }
 
-  def setStepSize(s: Double) = {
+  def setStepSize(s: Double): this.type = {
     this.step = s
     this
   }

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/optimization/ProbGPCommMachine.scala

@@ -181,7 +181,7 @@ object ProbGPCommMachine {
 
   def calculateModelWeightsSigmoid(
     baselineMethod: String = "mean")(
-    energyLandscape: List[(Double, Map[String, Double])])
+    energyLandscape: Seq[(Double, Map[String, Double])])
   : Seq[(Double, Map[String, Double])] = {
 
     val h = DenseVector(energyLandscape.map(_._1).toArray)

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/optimization/ProbGPMixtureMachine.scala

@@ -22,14 +22,13 @@ class ProbGPMixtureMachine[T, I: ClassTag](
   MixtureMachine[T, I, Double, PartitionedVector, PartitionedPSDMatrix, BlockedMultiVariateGaussian,
     MultGaussianPRV, AbstractGPRegressionModel[T, I]](model) {
 
-
   val (kernelPipe, noisePipe) = (system.covariance.asPipe, system.noiseModel.asPipe)
 
   def blockedHypParams = system.covariance.blocked_hyper_parameters ++ system.noiseModel.blocked_hyper_parameters
 
   def blockedState = system._current_state.filterKeys(blockedHypParams.contains)
 
-  implicit val transform: DataPipe[T, Seq[(I, Double)]] = DataPipe(system.dataAsSeq _)
+  implicit val transform: DataPipe[T, Seq[(I, Double)]] = DataPipe(system.dataAsSeq)
 
   override val confToModel = DataPipe((model_state: Map[String, Double]) =>
     AbstractGPRegressionModel(
@@ -39,5 +38,5 @@ class ProbGPMixtureMachine[T, I: ClassTag](
   override val mixturePipe = DataPipe2(
     (models: Seq[AbstractGPRegressionModel[T, I]], weights: DenseVector[Double]) =>
     StochasticProcessMixtureModel[T, I](models, weights))
-  
+
 }

scripts/stochasticPriors.sc

@@ -33,7 +33,7 @@ val gsmKernel = GaussianSpectralKernel[Double](3.5, 2.0, encoder)
 val n = new MAKernel(0.8)
 
 val gp_prior = new LinearTrendGaussianPrior[Double](gsmKernel, n, trendEncoder, 0.0, 0.0)
-//gp_prior.hyperPrior_(hyp_prior)
+gp_prior.hyperPrior_(hyp_prior)
 
 val sgp_prior = new LinearTrendESGPrior[Double](rbfc, n, trendEncoder, 0.75, 0.1, 0.0, 0.0)
 sgp_prior.hyperPrior_(sgp_hyp_prior)
@@ -63,13 +63,9 @@ val sgpModel = sgp_prior.posteriorModel(dataset)
 gp_prior.globalOptConfig_(Map("gridStep" -> "0.0", "gridSize" -> "1", "globalOpt" -> "GS", "policy" -> "GS"))
 val gpModel1 = gp_prior.posteriorModel(dataset)
 
-val mixt_machine = new ProbGPMixtureMachine(gpModel1)
-  .setGridSize(2)
-  .setStepSize(0.50)
-  .setLogScale(true)
-  .setMaxIterations(16)
+val mixt_machine = new ProbGPMixtureMachine(gpModel1).setPrior(hyp_prior).setGridSize(2).setStepSize(0.50).setLogScale(true).setMaxIterations(200).setNumSamples(3)
 
-val (mix_model, mixt_model_conf) = mixt_machine.optimize(gp_prior.covariance.state ++ gp_prior.noiseCovariance.state)
+val (mix_model, mixt_model_conf) = mixt_machine.optimize(gp_prior.covariance.effective_state ++ gp_prior.noiseCovariance.effective_state)
 
 
 val zs: MultGaussianPRV = gpModel.predictiveDistribution(xs)