Added coregionalization component for Cauchy Kernel, added multivariate gaussian scaling of attributes

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.28"
+val mainVersion = "v1.4-beta.29"
 
 val dataDirectory = settingKey[File]("The directory holding the data files for running example scripts")
 

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

@@ -25,7 +25,7 @@ import io.github.mandar2812.dynaml.models.ParameterizedLearner
 import io.github.mandar2812.dynaml.models.gp.AbstractGPRegressionModel
 import io.github.mandar2812.dynaml.optimization.{CoupledSimulatedAnnealing, GPMLOptimizer, GloballyOptWithGrad, GridSearch}
 import io.github.mandar2812.dynaml.pipes.{DataPipe, ReversibleScaler, Scaler, StreamDataPipe}
-import io.github.mandar2812.dynaml.utils.{GaussianScaler, MinMaxScaler}
+import io.github.mandar2812.dynaml.utils.{GaussianScaler, MinMaxScaler, MVGaussianScaler}
 import org.apache.log4j.Logger
 
 /**
@@ -340,6 +340,33 @@ object DynaMLPipe {
       (scaler(trainTest), (featuresScaler, targetsScaler))
     })
 
+  /**
+    * Scale a data set which is stored as a [[Stream]],
+    * return the scaled data as well as a [[MVGaussianScaler]] instance
+    * which can be used to reverse the scaled values to the original
+    * data.
+    * */
+  val multivariateGaussianScaling =
+  DataPipe((trainTest: Stream[(DenseVector[Double], DenseVector[Double])]) => {
+
+    val (num_features, num_targets) = (trainTest.head._1.length, trainTest.head._2.length)
+
+    val (m, sigma) = utils.getStatsMult(trainTest.map(tup =>
+      DenseVector(tup._1.toArray ++ tup._2.toArray)).toList)
+
+    val featuresScaler = new MVGaussianScaler(
+      m(0 until num_features),
+      sigma(0 until num_features, 0 until num_features))
+
+    val targetsScaler = new MVGaussianScaler(
+      m(num_features until num_features + num_targets),
+      sigma(num_features until num_features + num_targets, num_features until num_features + num_targets))
+
+    val scaler: ReversibleScaler[(DenseVector[Double], DenseVector[Double])] = featuresScaler * targetsScaler
+
+    (scaler(trainTest), (featuresScaler, targetsScaler))
+  })
+
 
   /**
     * Perform gaussian normalization on a data stream which
@@ -371,6 +398,36 @@ object DynaMLPipe {
       (scaler(trainTest._1), scaler(trainTest._2), (featuresScaler, targetsScaler))
     })
 
+  /**
+    * Scale a data set which is stored as a [[Stream]],
+    * return the scaled data as well as a [[MVGaussianScaler]] instance
+    * which can be used to reverse the scaled values to the original
+    * data.
+    * */
+  val multivariateGaussianScalingTrainTest =
+  DataPipe((trainTest: (Stream[(DenseVector[Double], DenseVector[Double])],
+    Stream[(DenseVector[Double], DenseVector[Double])])) => {
+
+    val (num_features, num_targets) = (trainTest._1.head._1.length, trainTest._1.head._2.length)
+
+    val (m, sigma) = utils.getStatsMult(trainTest._1.map(tup =>
+      DenseVector(tup._1.toArray ++ tup._2.toArray)).toList)
+
+    val featuresScaler = new MVGaussianScaler(
+      m(0 until num_features),
+      sigma(0 until num_features, 0 until num_features))
+
+    val targetsScaler = new MVGaussianScaler(
+      m(num_features until num_features + num_targets),
+      sigma(num_features until num_features + num_targets, num_features until num_features + num_targets))
+
+    val scaler: ReversibleScaler[(DenseVector[Double], DenseVector[Double])] = featuresScaler * targetsScaler
+
+    (scaler(trainTest._1), scaler(trainTest._2), (featuresScaler, targetsScaler))
+
+  })
+
+
   /**
     * Scale a data set which is stored as a [[Stream]],
     * return the scaled data as well as a [[MinMaxScaler]] instance

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

@@ -42,4 +42,18 @@ class CauchyCovFunc(private var sigma: Double)
   override def gradient(x: Double, y: Double): Map[String, Double] = {
     Map("sigma" -> 2.0*math.pow(evaluate(x,y),2)*math.pow(x-y, 2)/math.pow(state("sigma"), 3))
   }
-}
+}
+
+class CoRegCauchyKernel(bandwidth: Double) extends LocalSVMKernel[Int] {
+
+  override val hyper_parameters: List[String] = List("coRegSigma")
+
+  state = Map("coRegLB" -> bandwidth)
+
+  override def gradient(x: Int, y: Int): Map[String, Double] =
+    Map("CoRegSigma" -> 2.0*math.pow(evaluate(x,y), 2)*math.pow(x-y, 2)/math.pow(state("CoRegSigma"), 3))
+
+  override def evaluate(x: Int, y: Int): Double = {
+    1/(1 + math.pow(x-y, 2)/math.pow(state("coRegSigma"), 2))
+  }
+}