Modifed :*: operator for kernels and random variables to :*

Author: mandar2812

build.sbt

@@ -12,7 +12,7 @@ packageDescription := "DynaML is a scala library/repl for implementing and worki
   "which can be extended easily to implement advanced models for small and large scale applications.\n\n"+
   "But the library can also be used as an educational/research tool for data analysis."
 
-val mainVersion = "v1.4-beta.27"
+val mainVersion = "v1.4-beta.28"
 
 val dataDirectory = settingKey[File]("The directory holding the data files for running example scripts")
 

dynaml-core/scripts/probModels.scala

@@ -9,13 +9,13 @@ import spire.implicits._
 val n = RandomVariable(Gaussian(0.0, 0.25))
 val n1 = RandomVariable(Laplace(0.0, 0.25))
 
-val combinedDist = n :*: n1
+val combinedDist = n :* n1
 
 val cond2 = DataPipe((xy: (Double, Double)) => RandomVariable(new Gumbel(0.4*xy._1 + 0.6*xy._2, math.sin(xy._1+xy._2)/2.0)))
 
 val mod = ProbabilityModel(combinedDist, cond2)
 
-val combinedDist2 = n1 :*: n
+val combinedDist2 = n1 :* n
 
 val mod1 = ProbabilityModel(combinedDist2, cond2)
 
@@ -43,7 +43,7 @@ val priorMean = RandomVariable(new Gaussian(0.0, 1.0))
 
 val priorSigma = RandomVariable(new Gamma(7.5, 1.0))
 
-val prior = priorSigma :*: priorMean
+val prior = priorMean :* priorSigma
 
 val iidPrior = IIDRandomVarDistr(prior) _
 

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/kernels/DiracKernel.scala

@@ -63,4 +63,12 @@ class MAKernel(private var noiseLevel: Double = 1.0)
   : KernelMatrix[DenseMatrix[Double]] =
     new SVMKernelMatrix(DenseMatrix.eye[Double](length)*state("noiseLevel"), length)
 
-}
+}
+
+class CoRegDiracKernel extends LocalSVMKernel[Int] {
+  override val hyper_parameters: List[String] = List()
+
+  override def gradient(x: Int, y: Int): Map[String, Double] = Map()
+
+  override def evaluate(x: Int, y: Int): Double = if(x == y) 1.0 else 0.0
+}

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/kernels/Kernel.scala

@@ -177,7 +177,7 @@ CovarianceFunction[Index, Double, DenseMatrix[Double]] {
     }
   }
 
-  def :*:[T1](otherKernel: LocalScalarKernel[T1]): CompositeCovariance[(Index, T1)] = {
+  def :*[T1](otherKernel: LocalScalarKernel[T1]): CompositeCovariance[(Index, T1)] = {
 
     val firstkernel = this
 

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/kernels/RBFKernel.scala

@@ -195,3 +195,17 @@ class SECovFunc(private var band: Double = 1.0, private var h: Double = 2.0)
     Map("amplitude" -> 2.0*state("amplitude")*super.evaluate(x,y)) ++ super.gradient(x,y)
 
 }
+
+class CoRegRBFKernel(bandwidth: Double) extends LocalSVMKernel[Int] {
+
+  override val hyper_parameters: List[String] = List("coRegB")
+
+  state = Map("coRegB" -> bandwidth)
+
+  override def gradient(x: Int, y: Int): Map[String, Double] =
+    Map("coRegB" -> 1.0*evaluate(x,y)*math.pow(x-y,2)/math.pow(math.abs(state("coRegB")), 3))
+
+  override def evaluate(x: Int, y: Int): Double = {
+    math.exp(-1.0*math.pow(x-y, 2.0)/state("coRegB"))
+  }
+}

dynaml-core/src/main/scala-2.11/io/github/mandar2812/dynaml/probability/RandomVariable.scala

@@ -36,7 +36,7 @@ abstract class RandomVariable[Domain] {
     *              component of the cartesian product.
     *
     * */
-  def :*:[Domain1](other: RandomVariable[Domain1]): RandomVariable[(Domain, Domain1)] = {
+  def :*[Domain1](other: RandomVariable[Domain1]): RandomVariable[(Domain, Domain1)] = {
     val sam = this.sample
     RandomVariable(BifurcationPipe(sam,other.sample))
   }
@@ -163,7 +163,7 @@ trait RandomVarWithDistr[Domain, Dist <: Density[Domain]]
 
   override val underlyingDist: Dist
 
-  def :*:[Domain1, Dist1 <: Density[Domain1]](other: RandomVarWithDistr[Domain1, Dist1]):
+  def :*[Domain1, Dist1 <: Density[Domain1]](other: RandomVarWithDistr[Domain1, Dist1]):
   RandomVarWithDistr[(Domain, Domain1), Density[(Domain, Domain1)]] = {
     val sam = this.sample
     val dist = this.underlyingDist
@@ -190,7 +190,7 @@ abstract class ContinuousDistrRV[Domain](implicit ev: Field[Domain])
 
   override val sample = DataPipe(() => underlyingDist.sample())
 
-  def :*:[Domain1](other: ContinuousDistrRV[Domain1])(implicit ev1: Field[Domain1])
+  def :*[Domain1](other: ContinuousDistrRV[Domain1])(implicit ev1: Field[Domain1])
   : ContinuousDistrRV[(Domain, Domain1)] = {
     val sam = this.underlyingDist.draw _
     val dist = this.underlyingDist
@@ -225,7 +225,7 @@ abstract class DiscreteDistrRV[Domain]
 
   override val sample = DataPipe(() => underlyingDist.sample())
 
-  def :*:[Domain1](other: DiscreteDistrRV[Domain1]): DiscreteDistrRV[(Domain, Domain1)] = {
+  def :*[Domain1](other: DiscreteDistrRV[Domain1]): DiscreteDistrRV[(Domain, Domain1)] = {
 
     val dist = this.underlyingDist
     val sam = this.underlyingDist.draw _