Decomposable Kernels feature now working

To Do:
1. Re-implement addition and multiplication of kernels in terms of
DecomposableKernel class

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 = "v1.4.1-beta.6"
+val mainVersion = "v1.4.1-beta.7"
 
 val dataDirectory = settingKey[File]("The directory holding the data files for running example scripts")
 

conf/DynaMLInit.scala

@@ -35,6 +35,8 @@ import io.github.mandar2812.dynaml.examples._
 import io.github.mandar2812.dynaml.models.neuralnets.TransferFunctions._
 //The probability API
 import io.github.mandar2812.dynaml.probability._
+//Wavelet API
+import io.github.mandar2812.dynaml.wavelets._
 //OpenML support
 import io.github.mandar2812.dynaml.openml.OpenML
 //Renjin imports

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

@@ -24,13 +24,15 @@ import io.github.mandar2812.dynaml.evaluation.RegressionMetrics
 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.pipes._
 import io.github.mandar2812.dynaml.utils.{GaussianScaler, MVGaussianScaler, MinMaxScaler}
 import io.github.mandar2812.dynaml.wavelets.{GroupedHaarWaveletFilter, HaarWaveletFilter, InvGroupedHaarWaveletFilter, InverseHaarWaveletFilter}
 import org.apache.log4j.Logger
 import org.renjin.script.RenjinScriptEngine
 import org.renjin.sexp._
 
+import scala.reflect.ClassTag
+
 /**
   * @author mandar2812 datum 3/2/16.
   *
@@ -557,6 +559,34 @@ object DynaMLPipe {
 
   val invGroupedHaarWaveletFilter = (orders: Array[Int]) => InvGroupedHaarWaveletFilter(orders)
 
+
+  def genericReplicationEncoder[I](n: Int)(implicit tag: ClassTag[I]): Encoder[I, Array[I]] = Encoder[I, Array[I]]((v: I) => {
+    Array.fill[I](n)(v)
+  }, (vs: Array[I]) => {
+    vs.head
+  })
+
+  /**
+    * Creates an [[Encoder]] which can split
+    * [[DenseVector]] instances into uniform splits and
+    * put them back together.
+    * */
+  val breezeDVSplitEncoder = (n: Int) => Encoder((v: DenseVector[Double]) => {
+    v.toArray.grouped(n).map(DenseVector(_)).toArray
+  }, (vs: Array[DenseVector[Double]]) => {
+    DenseVector(vs.map(_.toArray).reduceLeft((a,b) => a++b))
+  })
+
+  /**
+    * Creates an [[Encoder]] which replicates a
+    * [[DenseVector]] instance n times.
+    * */
+  val breezeDVReplicationEncoder = (n: Int) => Encoder((v: DenseVector[Double]) => {
+    Array.fill(n)(v)
+  }, (vs: Array[DenseVector[Double]]) => {
+   vs.head
+  })
+
   def trainParametricModel[
   G, T, Q, R, S, M <: ParameterizedLearner[G, T, Q, R, S]
   ](regParameter: Double, step: Double = 0.05,

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

@@ -1,6 +1,7 @@
 package io.github.mandar2812.dynaml.kernels
 
 import breeze.linalg.DenseMatrix
+import io.github.mandar2812.dynaml.DynaMLPipe
 import io.github.mandar2812.dynaml.algebra.{PartitionedPSDMatrix, PartitionedVector}
 import io.github.mandar2812.dynaml.pipes.{DataPipe, Encoder}
 
@@ -82,25 +83,35 @@ abstract class CompositeCovariance[T]
   */
 class DecomposableCovariance[S](kernels: LocalScalarKernel[S]*)(
   implicit encoding: Encoder[S, Array[S]],
-  reducer: DataPipe[Array[Double], Double]) extends CompositeCovariance[S] {
+  reducer: DataPipe[Array[Double], Double] = DecomposableCovariance.:+:) extends CompositeCovariance[S] {
 
-  val kernelMap = kernels.map(k => (k.toString.split(".").last, k)).toMap
+  val kernelMap = kernels.map(k => (k.toString.split("\\.").last, k)).toMap
+
+  state = kernels.map(k => {
+    val id = k.toString.split("\\.").last
+    k.state.map(h => (id+"/"+h._1, h._2))
+  }).reduceLeft(_++_)
 
   override val hyper_parameters: List[String] = kernels.map(k => {
-    val id = k.toString.split(".").last
+    val id = k.toString.split("\\.").last
     k.hyper_parameters.map(h => id+"/"+h)
   }).reduceLeft(_++_)
 
   blocked_hyper_parameters = kernels.map(k => {
-    val id = k.toString.split(".").last
+    val id = k.toString.split("\\.").last
     k.blocked_hyper_parameters.map(h => id+"/"+h)
   }).reduceLeft(_++_)
 
+  override def repr: DecomposableCovariance[S] = this
+
   override def setHyperParameters(h: Map[String, Double]): DecomposableCovariance.this.type = {
+    //Sanity Check
+    assert(effective_hyper_parameters.forall(h.contains),
+      "All hyper parameters must be contained in the arguments")
     //group the hyper params by kernel id
     h.toSeq.map(kv => {
       val idS = kv._1.split("/")
-      (idS.head, (idS.last, kv._2))
+      (idS.head, (idS.last.mkString("/"), kv._2))
     }).groupBy(_._1).map(hypC => {
       val kid = hypC._1
       val hyper_params = hypC._2.map(_._2).toMap
@@ -116,4 +127,12 @@ class DecomposableCovariance[S](kernels: LocalScalarKernel[S]*)(
       coupleAndKern._2.evaluate(u,v)
     }))
   }
+}
+
+object DecomposableCovariance {
+
+  val :+: = DataPipe((l: Array[Double]) => l.sum)
+
+  val :*: = DataPipe((l: Array[Double]) => l.sum)
+
 }

dynaml-pipes/src/main/scala-2.11/io/github/mandar2812/dynaml/pipes/Scaler.scala

@@ -40,13 +40,13 @@ object Scaler {
   *
   *
   * */
-trait ReversibleScaler[S] extends Scaler[S] {
+trait ReversibleScaler[S] extends Scaler[S] with Encoder[S, S]{
 
   /**
     * The inverse operation of this scaling.
     *
     * */
-  val i: Scaler[S]
+  override val i: Scaler[S]
 
   def *[T](that: ReversibleScaler[T]) = {