Code clean up.

build.sbt

@@ -16,10 +16,16 @@ lazy val root = (project in file(".")).
     libraryDependencies ++= Seq(
       "org.apache.commons" % "commons-math3" % "3.3",
       "com.typesafe.scala-logging" %% "scala-logging-slf4j" % "2.1.2",
-      "org.scalanlp" %% "breeze" % "0.11.2",
-      "org.scalanlp" %% "breeze-natives" % "0.11.2",
+      "org.scalanlp" %% "breeze" % "0.12-SNAPSHOT",
       // test scoped
       "org.slf4j" % "slf4j-log4j12" % "1.7.2" % Test,
       "com.novocode" % "junit-interface" % "0.11" % Test
+    ),
+
+    resolvers ++= Seq(
+    // other resolvers here
+    // if you want to use snapshot builds (currently 0.12-SNAPSHOT), use this.
+    "Sonatype Snapshots" at "https://oss.sonatype.org/content/repositories/snapshots/",
+    "Sonatype Releases" at "https://oss.sonatype.org/content/repositories/releases/"
     )
   )

doc/lowlevel/README.md

@@ -32,18 +32,25 @@ None
 *   Continuous Bayesian Networks
     * [Linear Gaussian Models - 1D localisation, 4 approaches: Canonical Gaussian, Bayes's theorem for Gaussian Variables, Expectation Propagation and Kalman Filter](https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/localisation_example/localisation_example.md)
 *   Hybrid Bayesian Networks
-    * [Gaussian approximation with moment matching, aka proj() operator in Expectation Propagation](https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/moment_matching/moment_matching.md)
-    * [Expectation Propagation for the Clutter Problem](https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/clutter_problem_ep/clutter_problem_ep.md)
-    * [TrueSkill - Updating player skills in tennis with Expectation Propagation inference algorithm](https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/trueskill_in_tennis/trueskill_in_tennis.md)
-    * [TrueSkill on a factor graph in Tennis] (https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/trueskill_in_tennis_factor_graph/trueskill_in_tennis_factor_graph.md)
-    * [TrueSkill on a factor graph in Tennis (Dynamic Bayesian Network for 3 players with 6 games over 3 time slices)] (https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/trueskill_in_tennis_factor_graph_dbn/trueskill_in_tennis_factor_graph_dbn.md)
+	* EP   
+    	* [Gaussian approximation with moment matching, aka proj() operator in Expectation Propagation](https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/moment_matching/moment_matching.md)
+    	* [Expectation Propagation for the Clutter Problem](https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/clutter_problem_ep/clutter_problem_ep.md)
+	* TrueSkill
+    	* [TrueSkill - Updating player skills in tennis with Expectation Propagation inference algorithm](https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/trueskill_in_tennis/trueskill_in_tennis.md)
+    	* [TrueSkill on a factor graph in Tennis] (https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/trueskill_in_tennis_factor_graph/trueskill_in_tennis_factor_graph.md)
+    	* [TrueSkill on a factor graph in Tennis (Dynamic Bayesian Network for 3 players with 6 games over 3 time slices)] (https://github.com/danielkorzekwa/bayes-scala/blob/master/doc/trueskill_in_tennis_factor_graph_dbn/trueskill_in_tennis_factor_graph_dbn.md)
 * Code examples only
-    * [Linear Dynamical Systems M-step (prior mean, emission variance, transition variance)](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/learn/lds/GenericLDSMStepTest.scala)
-    * [Linear Dynamical Systems EM (learning prior mean and emission variance only from multiple data sequences)](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/learn/lds/GenericLDSEMTest.scala)
-    * [Inference in Gaussian Process regression](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/infer/gp/gpr/GenericGPRegressionTest.scala)
-    * [Inference in Gaussian Process regression with inducing variables - link to bayes-scala-gp project](https://github.com/danielkorzekwa/bayes-scala-gp/blob/master/src/test/scala/dk/gp/sgpr/sgprPredictTest.scala)
-    * [Parameters learning in Gaussian Process regression with inducing variables (variational lower bound on marginal likelihood) - link to bayes-scala-gp project] (https://github.com/danielkorzekwa/bayes-scala-gp/blob/master/src/test/scala/dk/gp/sgpr/sgprTrainTest.scala)
-    * [Parameters learning in Gaussian Process regression (marginal likelihood maximisation)](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/infer/gp/gpr/GpmlRegressionLearnTest.scala)
+	* Linear Dynamical Systems 	
+    	* [Linear Dynamical Systems M-step (prior mean, emission variance, transition variance)](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/learn/lds/GenericLDSMStepTest.scala)
+    	* [Linear Dynamical Systems EM (learning prior mean and emission variance only from multiple data sequences)](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/learn/lds/GenericLDSEMTest.scala)
+    * Gaussian Process Regression
+    	* [Inference in Gaussian Process regression](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/infer/gp/gpr/GenericGPRegressionTest.scala)
+    	* [Parameters learning in Gaussian Process regression (marginal likelihood maximisation)](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/infer/gp/gpr/GpmlRegressionLearnTest.scala)
+	* Sparse Gaussian Process Regression
+    	* [Inference in Gaussian Process regression with inducing variables - link to bayes-scala-gp project](https://github.com/danielkorzekwa/bayes-scala-gp/blob/master/src/test/scala/dk/gp/sgpr/sgprPredictTest.scala)
+    	* [Parameters learning in Gaussian Process regression with inducing variables (variational lower bound on marginal likelihood) - link to bayes-scala-gp project] (https://github.com/danielkorzekwa/bayes-scala-gp/blob/master/src/test/scala/dk/gp/sgpr/sgprTrainTest.scala)
+    	* 
+
     * [Parameters learning in Gaussian Process regression  (variational lower bound on marginal likelihood)](https://github.com/danielkorzekwa/bayes-scala/blob/master/src/test/scala/dk/bayes/infer/gp/infercovparamsem/inferCovParamsEmTest.scala)
                                  
 *   Others 

project/build.properties

@@ -1 +1 @@
-sbt.version=0.13.5
+sbt.version=0.13.7

src/main/scala/dk/bayes/dsl/Variable.scala

@@ -22,11 +22,11 @@ trait Variable {
 
   init()
 
-  private def init() {
+  private def init() = {
     getParents().foreach(p => p.addChild(this))
   }
 
-  def addChild(v: Variable) {
+  def addChild(v: Variable) = {
     children += v
   }
   def getChildren(): Seq[Variable] = children.toList
@@ -40,7 +40,7 @@ trait Variable {
   def getAllVariables(): Seq[Variable] = {
     val variables = new HashSet[Variable]()
 
-    def addVariable(v: Variable) {
+    def addVariable(v: Variable):Unit = {
 
       if (!variables.contains(v)) {
         variables += v

src/main/scala/dk/bayes/dsl/variable/gaussian/multivariate/inferMultivariateGaussianSimplest.scala

@@ -22,7 +22,6 @@ object inferMultivariateGaussianSimplest extends InferEngine[MultivariateGaussia
 
     (child.getParents().size == 1 && child.getParents()(0).eq(x)) &&
       !child.hasChildren &&
-      child.b.size == x.m.size &&
       child.yValue.isDefined
 
   }

src/main/scala/dk/bayes/infer/ClusterGraphInfer.scala

@@ -18,7 +18,7 @@ trait ClusterGraphInfer {
    * @param messageOrder Order of clusters in which messages are sent for a single iteration of Belief Propagation
    *
    */
-  def calibrate(iterNum: (Int) => Unit, messageOrder: MessageOrder)
+  def calibrate(iterNum: (Int) => Unit, messageOrder: MessageOrder):Unit
 
   /**
    * Applies evidence and calibrates cluster graph.
@@ -55,5 +55,5 @@ trait ClusterGraphInfer {
    *
    *  @param evidence Tuple2[variableId, variable value]
    */
-  def setEvidence(evidence: Tuple2[Int, Int])
+  def setEvidence(evidence: Tuple2[Int, Int]):Unit
 }

src/main/scala/dk/bayes/infer/LoopyBP.scala

@@ -21,7 +21,7 @@ import LoopyBP._
  */
 case class LoopyBP(clusterGraph: ClusterGraph, threshold: Double = 0.00001) extends ClusterGraphInfer {
 
-  def calibrate(iterNum: (Int) => Unit = (iterNum: Int) => {}, messageOrder: MessageOrder = ForwardBackwardMsgOrder()) {
+  def calibrate(iterNum: (Int) => Unit = (iterNum: Int) => {}, messageOrder: MessageOrder = ForwardBackwardMsgOrder()):Unit = {
 
     @tailrec
     def calibrateUntilConverge(currentIter: Int): ClusterGraph = {
@@ -58,7 +58,7 @@ case class LoopyBP(clusterGraph: ClusterGraph, threshold: Double = 0.00001) exte
     evidenceLogLikelihood
   }
 
-  private def calibrateCluster(cluster: Cluster) {
+  private def calibrateCluster(cluster: Cluster):Unit = {
 
     cluster.getEdges().foreach { edge =>
 
@@ -137,7 +137,7 @@ case class LoopyBP(clusterGraph: ClusterGraph, threshold: Double = 0.00001) exte
     marginalFactor
   }
 
-  def setEvidence(evidence: Tuple2[Int, Int]) {
+  def setEvidence(evidence: Tuple2[Int, Int]):Unit = {
 
     for (cluster <- clusterGraph.getClusters()) {
 

src/main/scala/dk/bayes/infer/ep/EP.scala

@@ -20,7 +20,7 @@ trait EP {
    * @param varId  Variable id
    * @param varValue Variable value
    */
-  def setEvidence(varId: Int, varValue: AnyVal)
+  def setEvidence(varId: Int, varValue: AnyVal):Unit
 
   /**
    * Returns marginal factor for a given variable(s) in a factor graph.

src/main/scala/dk/bayes/infer/ep/calibrate/fb/ForwardBackwardEPCalibrate.scala

@@ -66,7 +66,7 @@ case class ForwardBackwardEPCalibrate(factorGraph: FactorGraph, threshold: Doubl
    * Executes a single message passing routine on a factor graph.
    *
    */
-  private def calibrateIteration(nodes: Seq[Node], newMsgIndex: () => Long) {
+  private def calibrateIteration(nodes: Seq[Node], newMsgIndex: () => Long):Unit = {
 
     val nodesNum = nodes.size
 
@@ -80,7 +80,7 @@ case class ForwardBackwardEPCalibrate(factorGraph: FactorGraph, threshold: Doubl
     }
   }
 
-  private def sendFactorMessage(factorNode: FactorNode, newMsgIndex: () => Long) {
+  private def sendFactorMessage(factorNode: FactorNode, newMsgIndex: () => Long):Unit = {
 
     factorNode match {
       case factorNode: SingleFactorNode => sendFactorMessage(factorNode, newMsgIndex)
@@ -91,7 +91,7 @@ case class ForwardBackwardEPCalibrate(factorGraph: FactorGraph, threshold: Doubl
 
   }
 
-  private def sendFactorMessage(factorNode: SingleFactorNode, newMsgIndex: () => Long) {
+  private def sendFactorMessage(factorNode: SingleFactorNode, newMsgIndex: () => Long):Unit = {
     val newMessage = factorNode.getFactor().asInstanceOf[SingleFactor]
 
     factorNode.gate.setMessage(newMessage, newMsgIndex())
@@ -100,7 +100,7 @@ case class ForwardBackwardEPCalibrate(factorGraph: FactorGraph, threshold: Doubl
 
   }
 
-  private def sendFactorMessage(factorNode: DoubleFactorNode, newMsgIndex: () => Long) {
+  private def sendFactorMessage(factorNode: DoubleFactorNode, newMsgIndex: () => Long):Unit = {
 
     val gate1MsgIn = factorNode.gate1.getEndGate.getMessage
     val gate2MsgIn = factorNode.gate2.getEndGate.getMessage
@@ -116,7 +116,7 @@ case class ForwardBackwardEPCalibrate(factorGraph: FactorGraph, threshold: Doubl
 
   }
 
-  private def sendFactorMessage(factorNode: TripleFactorNode, newMsgIndex: () => Long) {
+  private def sendFactorMessage(factorNode: TripleFactorNode, newMsgIndex: () => Long):Unit = {
 
     val gate1MsgIn = factorNode.gate1.getEndGate.getMessage
     val gate2MsgIn = factorNode.gate2.getEndGate.getMessage
@@ -136,7 +136,7 @@ case class ForwardBackwardEPCalibrate(factorGraph: FactorGraph, threshold: Doubl
 
   }
 
-  private def sendFactorMessage(factorNode: GenericFactorNode, newMsgIndex: () => Long) {
+  private def sendFactorMessage(factorNode: GenericFactorNode, newMsgIndex: () => Long):Unit = {
     val msgsIn = factorNode.gates.map(g => g.getEndGate.getMessage())
     val factor = factorNode.getFactor().asInstanceOf[GenericFactor]
 
@@ -152,7 +152,7 @@ case class ForwardBackwardEPCalibrate(factorGraph: FactorGraph, threshold: Doubl
   }
 
   /**Returns the number of messages sent.*/
-  private def sendVariableMessage(varNode: VarNode, newMsgIndex: () => Long) {
+  private def sendVariableMessage(varNode: VarNode, newMsgIndex: () => Long):Unit = {
 
     var marginalFactor = varNode.getGates()(0).getEndGate.getMessage()
     var i = 1

src/main/scala/dk/bayes/infer/epnaivebayes/EPNaiveBayesFactorGraph.scala

@@ -28,10 +28,10 @@ case class EPNaiveBayesFactorGraph[X](prior: SingleFactor[X], likelihoods: Seq[D
 
   def getPosterior(): X = posterior
 
-  def calibrate(maxIter: Int = 100, threshold: Double = 1e-6) {
+  def calibrate(maxIter: Int = 100, threshold: Double = 1e-6):Unit = {
 
     @tailrec
-    def calibrateIter(currPosterior: X, iterNum: Int) {
+    def calibrateIter(currPosterior: X, iterNum: Int):Unit = {
       if (iterNum >= maxIter) {      
         logger.warn(s"Factor graph did not converge in less than ${maxIter} iterations. Prior=%s, Posterior=%s".format(prior, posterior))
         return
@@ -45,7 +45,7 @@ case class EPNaiveBayesFactorGraph[X](prior: SingleFactor[X], likelihoods: Seq[D
     calibrateIter(posterior, 1)
   }
 
-  private def sendMsgsParallel() {
+  private def sendMsgsParallel():Unit = {
 
     msgsUp = msgsUp.zip(likelihoods).map {
       case (currMsgUp, llh) =>
@@ -61,7 +61,7 @@ case class EPNaiveBayesFactorGraph[X](prior: SingleFactor[X], likelihoods: Seq[D
     posterior = multOp(prior.factorMsgDown, multOp(msgsUp: _*))
   }
 
-  private def sendMsgsSerial() {
+  private def sendMsgsSerial():Unit = {
 
     msgsUp = msgsUp.zip(likelihoods).map {
       case (currMsgUp, llh) =>

src/main/scala/dk/bayes/learn/em/EMLearn.scala

@@ -31,7 +31,7 @@ trait EMLearn {
    *
    *  @param progress Progress monitoring. It is called by this method at the end of every iteration
    */
-  def learn(clusterGraph: ClusterGraph, trainSet: DataSet, maxIterNum: Int, progress: (Progress) => Unit)
+  def learn(clusterGraph: ClusterGraph, trainSet: DataSet, maxIterNum: Int, progress: (Progress) => Unit):Unit
 }
 
 object EMLearn {

src/main/scala/dk/bayes/learn/em/GenericEMLearn.scala

@@ -50,7 +50,7 @@ object GenericEMLearn extends EMLearn {
 
   }
 
-  private def updateInitialClusterPotentials(clusterGraph: ClusterGraph, clusterPotentialsByTypeId: Map[Int, Factor]) {
+  private def updateInitialClusterPotentials(clusterGraph: ClusterGraph, clusterPotentialsByTypeId: Map[Int, Factor]):Unit = {
     for (cluster <- clusterGraph.getClusters()) {
       val clusterTypePotentials = clusterPotentialsByTypeId(cluster.typeId)
       val newClusterPotentials = cluster.getFactor().copy(clusterTypePotentials.getValues())

src/main/scala/dk/bayes/model/clustergraph/Cluster.scala

@@ -1,6 +1,6 @@
 package dk.bayes.model.clustergraph
 
-import factor._
+import factor.Factor
 
 /**
  * Represents cluster in a cluster graph.
@@ -20,19 +20,19 @@ class Cluster(val id: Int, val typeId: Int, factor: Factor) {
   private var _factor: Factor = factor
   private var edges: List[Edge] = List()
 
-  def addEdge(edge: Edge) { edges = edge :: edges }
+  def addEdge(edge: Edge):Unit = { edges = edge :: edges }
 
   def getEdges(): Seq[Edge] = edges
 
   def getFactor(): Factor = _factor
 
-  def updateFactor(newFactor: Factor) {
+  def updateFactor(newFactor: Factor):Unit = {
     _factor = newFactor
 
     resetMessages()
   }
 
-  def resetMessages() {
+  def resetMessages():Unit = {
     edges.foreach { edge => edge.resetMessage() }
   }
 }

src/main/scala/dk/bayes/model/clustergraph/ClusterGraph.scala

@@ -22,20 +22,20 @@ trait ClusterGraph {
    *
    * @param factor Initial cluster potentials
    */
-  def addCluster(clusterId: Int, factor: Factor, clusterTypeId: Option[Int]=None)
+  def addCluster(clusterId: Int, factor: Factor, clusterTypeId: Option[Int]=None):Unit
 
   /**
    * Adds edge between clusters in this cluster graph.
    */
-  def addEdge(clusterId1: Int, clusterId2: Int)
+  def addEdge(clusterId1: Int, clusterId2: Int):Unit
 
   /**
    * Adds edges between clusters in this cluster graph.
    *
    * @param firstEdge Tuple2[clusterId, clusterId2]
    * @param nextEdges Tuple2[clusterId, clusterId2]
    */
-  def addEdges(firstEdge: Tuple2[Int, Int], nextEdges: Tuple2[Int, Int]*)
+  def addEdges(firstEdge: Tuple2[Int, Int], nextEdges: Tuple2[Int, Int]*):Unit
 
   /**
    * Returns all clusters in this cluster graph.

src/main/scala/dk/bayes/model/clustergraph/Edge.scala

@@ -17,17 +17,17 @@ class Edge(val destClusterId: Int, val sepsetVariable: Var) {
   private var newMessage: SingleFactor = SingleFactor(sepsetVariable, Array.fill(sepsetVariable.dim)(1d))
   private var oldMessage: SingleFactor = newMessage
 
-  def setIncomingEdge(edge: Edge) {
+  def setIncomingEdge(edge: Edge):Unit = {
     incomingEdge = Some(edge)
   }
   def getIncomingEdge(): Option[Edge] = incomingEdge
 
-  def resetMessage() {
+  def resetMessage():Unit = {
     newMessage = SingleFactor(sepsetVariable, Array.fill(sepsetVariable.dim)(1d))
     oldMessage = newMessage
   }
 
-  def updateMessage(message: SingleFactor) {
+  def updateMessage(message: SingleFactor):Unit = {
     oldMessage = newMessage
     newMessage = message
   }

src/main/scala/dk/bayes/model/clustergraph/GenericClusterGraph.scala

@@ -27,7 +27,7 @@ case class GenericClusterGraph() extends ClusterGraph {
 
   def getCluster(clusterId: Int): Cluster = clusters.find(c => c.id == clusterId).get
 
-  def addEdge(clusterId1: Int, clusterId2: Int) {
+  def addEdge(clusterId1: Int, clusterId2: Int):Unit = {
     val cluster1 = clusters.find(c => c.id == clusterId1).get
     val cluster2 = clusters.find(c => c.id == clusterId2).get
     val sepsetVariable = calcSepsetVariable(cluster1, cluster2)
@@ -42,7 +42,7 @@ case class GenericClusterGraph() extends ClusterGraph {
     cluster2.addEdge(edge21)
   }
 
-  def addEdges(firstEdge: Tuple2[Int, Int], nextEdges: Tuple2[Int, Int]*) {
+  def addEdges(firstEdge: Tuple2[Int, Int], nextEdges: Tuple2[Int, Int]*):Unit = {
     addEdge(firstEdge._1, firstEdge._2)
     nextEdges.foreach(e => addEdge(e._1, e._2))
   }

src/main/scala/dk/bayes/model/clustergraph/factor/MultiFactor.scala

@@ -141,7 +141,7 @@ class MultiFactor(variables: Array[Var], values: Array[Double]) extends Factor {
    * @param stepSize Number of steps before reaching next assignment for a given dimension
    * @param process (valueIndex, variableIndex) => Unit
    */
-  private def processValues(dim: Int, stepSize: Int, process: (Int, Int) => Unit) {
+  private def processValues(dim: Int, stepSize: Int, process: (Int, Int) => Unit):Unit = {
 
     var i = 0
     var varIndex = 0

src/main/scala/dk/bayes/model/factorgraph/FactorGraph.scala

@@ -13,7 +13,7 @@ trait FactorGraph {
   /**
    * Adds factor to the factor graph.
    */
-  def addFactor(factor: Factor)
+  def addFactor(factor: Factor):Unit
 
   /**
    * Returns all nodes (factors and variables) in a graph.

src/main/scala/dk/bayes/model/factorgraph/Gate.scala

@@ -23,10 +23,10 @@ sealed abstract class Gate(initialMsg: SingleFactor) {
   /**Allows for comparing the age between different messages and finding the message that was updated least recently.*/
   private var msgIndex: Long = -1
 
-  def setEndGate(gate: END_GATE) { endGate = Some(gate) }
+  def setEndGate(gate: END_GATE):Unit = { endGate = Some(gate) }
   def getEndGate(): END_GATE = endGate.get
 
-  def setMessage(newMessage: SingleFactor, msgIndex: Long) {
+  def setMessage(newMessage: SingleFactor, msgIndex: Long):Unit = {
     oldMessage = message
     message = newMessage
 

src/main/scala/dk/bayes/model/factorgraph/Node.scala

@@ -32,7 +32,7 @@ sealed abstract class FactorNode(factor: Factor) extends Node {
   private var _factor: Factor = factor
 
   def getFactor(): Factor = _factor
-  def setFactor(factor: Factor) { _factor = factor }
+  def setFactor(factor: Factor):Unit = { _factor = factor }
 
   /**
    * Returns the product of the factor and all incoming messages.

src/test/scala/dk/bayes/dsl/demo/ClutterProblemTest.scala

@@ -13,7 +13,7 @@ import dk.bayes.dsl.infer
  */
 class ClutterProblemTest {
 
-  @Test def test {
+  @Test def test:Unit = {
 
     val x = Gaussian(15, 100)
     val y1 = ClutteredGaussian(x, w = 0.4, a = 10, value = 3)