Skip to content

Commit

Permalink
Uses DenseVector instead of Vector
Browse files Browse the repository at this point in the history
  • Loading branch information
@bethard
bethard committed May 22, 2013
1 parent f0d165d commit 361c8a8
Showing 1 changed file with 23 additions and 23 deletions.
46 changes: 23 additions & 23 deletions src/org/goobs/nlp/Convex.scala
View file
Expand Up @@ -4,16 +4,16 @@ import breeze.linalg._
import breeze.numerics._
import scala.util.Random

trait ObjectiveFn extends Function1[Vector[Double],Option[Double]] {
trait ObjectiveFn extends Function1[DenseVector[Double],Option[Double]] {
def cardinality:Int

def gradient(x:Vector[Double]):Option[Vector[Double]] = None
def hessian(x:Vector[Double]):Option[Matrix[Double]] = None
def gradient(x:DenseVector[Double]):Option[DenseVector[Double]] = None
def hessian(x:DenseVector[Double]):Option[Matrix[Double]] = None

def differentiableAt(x:Vector[Double]):Boolean = gradient(x).isDefined
def twiceDifferentiableAt(x:Vector[Double]):Boolean = hessian(x).isDefined
def differentiableAt(x:DenseVector[Double]):Boolean = gradient(x).isDefined
def twiceDifferentiableAt(x:DenseVector[Double]):Boolean = hessian(x).isDefined

def plot(x:Vector[Double],hold:Boolean=false){
def plot(x:DenseVector[Double],hold:Boolean=false){
if(cardinality > 1){
throw new IllegalStateException("Cannot plot function of cardinality > 1")
}
Expand All @@ -34,7 +34,7 @@ trait ObjectiveFn extends Function1[Vector[Double],Option[Double]] {
}
}

case class OptimizerProfile(optimalX:Vector[Double],optimalValue:Double,guessProfile:Vector[Double]) {
case class OptimizerProfile(optimalX:DenseVector[Double],optimalValue:Double,guessProfile:DenseVector[Double]) {
def plotObjective(name:String="objective"){
breeze.plot.plot(
x=DenseVector(Array.range(0,guessProfile.length).map(_.toDouble)),
Expand All @@ -59,7 +59,7 @@ case class OptimizerProfile(optimalX:Vector[Double],optimalValue:Double,guessPro
}

trait Optimizer {
def minimize(fn:ObjectiveFn, initialValue:Vector[Double]):OptimizerProfile
def minimize(fn:ObjectiveFn, initialValue:DenseVector[Double]):OptimizerProfile
}

object Optimizer {
Expand All @@ -78,10 +78,10 @@ abstract class DescentOptimizer(tolerance:Double, lineStep:Double) extends Optim
}


def converged(fnValue:Double,grad:Vector[Double],hessian:()=>Matrix[Double]):Boolean
def delta(fnValue:Double,gradient:Vector[Double],hessian:()=>Matrix[Double]):Vector[Double]
def converged(fnValue:Double,grad:DenseVector[Double],hessian:()=>Matrix[Double]):Boolean
def delta(fnValue:Double,gradient:DenseVector[Double],hessian:()=>Matrix[Double]):DenseVector[Double]

protected def safeMultiply(v:Vector[Double], t:Double):Vector[Double] = {
protected def safeMultiply(v:DenseVector[Double], t:Double):DenseVector[Double] = {
if(t == 0.0){
DenseVector.zeros[Double](v.length)
} else if(t < 1e-5){
Expand Down Expand Up @@ -119,11 +119,11 @@ abstract class DescentOptimizer(tolerance:Double, lineStep:Double) extends Optim
}
}

private def moveDeltaT(x:Vector[Double], delta:Vector[Double], t:Double):Vector[Double] = {
private def moveDeltaT(x:DenseVector[Double], delta:DenseVector[Double], t:Double):DenseVector[Double] = {
x :+ safeMultiply(delta,t)
}

private def lineSearch(fn:ObjectiveFn, x:Vector[Double], delta:Vector[Double], gradient:Vector[Double]):Double = {
private def lineSearch(fn:ObjectiveFn, x:DenseVector[Double], delta:DenseVector[Double], gradient:DenseVector[Double]):Double = {
var t:Double = 1.0
def check(t:Double):Boolean = {
fn(x).flatMap{ (fnValue:Double) =>
Expand All @@ -145,7 +145,7 @@ abstract class DescentOptimizer(tolerance:Double, lineStep:Double) extends Optim
t
}

def minimize(fn:ObjectiveFn, initialValue:Vector[Double]):OptimizerProfile = {
def minimize(fn:ObjectiveFn, initialValue:DenseVector[Double]):OptimizerProfile = {
//--Initialize
//(variables)
var x = initialValue
Expand Down Expand Up @@ -205,10 +205,10 @@ abstract class DescentOptimizer(tolerance:Double, lineStep:Double) extends Optim

class GradientDescentOptimizer(gradientTolerance:Double,tolerance:Double,lineStep:Double
) extends DescentOptimizer(tolerance, lineStep) {
override def converged(fnValue:Double,grad:Vector[Double],hessian:()=>Matrix[Double]):Boolean
override def converged(fnValue:Double,grad:DenseVector[Double],hessian:()=>Matrix[Double]):Boolean
= grad.forallValues{ _.abs <= gradientTolerance}

override def delta(fnValue:Double,grad:Vector[Double],hessian:()=>Matrix[Double]):Vector[Double]
override def delta(fnValue:Double,grad:DenseVector[Double],hessian:()=>Matrix[Double]):DenseVector[Double]
= -grad
}

Expand All @@ -227,10 +227,10 @@ class NewtonOptimizer(lambdaTolerance:Double,hessianInterval:Int,tolerance:Doubl
hessianInverseCache.get
}

private def lambdaSquared(grad:Vector[Double],hessian:()=>Matrix[Double]):Double = grad.t * inv(hessian) * grad
override def converged(fnValue:Double,grad:Vector[Double],hessian:()=>Matrix[Double]):Boolean
private def lambdaSquared(grad:DenseVector[Double],hessian:()=>Matrix[Double]):Double = grad.t * inv(hessian) * grad
override def converged(fnValue:Double,grad:DenseVector[Double],hessian:()=>Matrix[Double]):Boolean
= lambdaSquared(grad,hessian) / 1.0 <= lambdaTolerance
override def delta(fnValue:Double,grad:Vector[Double],hessian:()=>Matrix[Double]):Vector[Double]
override def delta(fnValue:Double,grad:DenseVector[Double],hessian:()=>Matrix[Double]):DenseVector[Double]
= -inv(hessian)*grad
}

Expand All @@ -255,7 +255,7 @@ object Convex {
//--Objective Function
val fn:ObjectiveFn = new ObjectiveFn {
def cardinality:Int = n
def apply(x: Vector[Double]):Option[Double] = {
def apply(x: DenseVector[Double]):Option[Double] = {
val value = (0 until m).map{ (i:Int) => log( 1 :- (A(i,::) * x) ) }.sum +
x.map{ (v) => log(1-v*v) }.sum
if(value.isNaN){
Expand All @@ -264,14 +264,14 @@ object Convex {
Some(-value)
}
}
override def gradient(x:Vector[Double]):Option[Vector[Double]] = {
override def gradient(x:DenseVector[Double]):Option[DenseVector[Double]] = {
apply(x).flatMap{ (fnValue:Double) =>
val termA = (0 until m).map{ (i:Int) =>
val numer = A(i,::).t
val denom = 1.0 - (A(i,::) * x)
numer :/ denom
}.foldLeft(DenseVector.zeros[Double](cardinality)){
case (soFar:DenseVector[Double], term:Vector[Double]) => soFar :+ term
case (soFar:DenseVector[Double], term:DenseVector[Double]) => soFar :+ term
}
val termB = (2 :* x) :/ ( (x :^ 2) :- 1.0)
val deriv = termA :- termB
Expand All @@ -282,7 +282,7 @@ object Convex {
}
}
}
override def hessian(x:Vector[Double]):Option[Matrix[Double]] = {
override def hessian(x:DenseVector[Double]):Option[Matrix[Double]] = {
val hessian:Matrix[Double] =
if(diagonal){
val hessian = DenseMatrix.eye[Double](x.length)
Expand Down

0 comments on commit 361c8a8

Please sign in to comment.