breeze.optimize.DiffFunction.scala Maven / Gradle / Ivy

Go to download
Show more of this group Show more artifacts with this name
Show all versions of breeze_2.10 Show documentation
breeze
There is a newer version: 0.9
package breeze.optimize

import breeze.linalg.support.CanCopy
import breeze.math.InnerProductModule
import breeze.util.Isomorphism

/*
 Copyright 2009 David Hall, Daniel Ramage
 
 Licensed under the Apache License, Version 2.0 (the "License")
 you may not use this file except in compliance with the License.
 You may obtain a copy of the License at 
 
 http://www.apache.org/licenses/LICENSE-2.0
 
 Unless required by applicable law or agreed to in writing, software
 distributed under the License is distributed on an "AS IS" BASIS,
 WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 See the License for the specific language governing permissions and
 limitations under the License. 
*/

/**
* Represents a differentiable function whose output is guaranteed to be consistent
*
* @author dlwh
*/
trait DiffFunction[T] extends StochasticDiffFunction[T] { outer =>
  def cached(implicit copy: CanCopy[T]) = {
    if (this.isInstanceOf[CachedDiffFunction[_]]) {
      this
    } else {
      new CachedDiffFunction(this)
    }
  }

  override def throughLens[U](implicit l: Isomorphism[T,U]):DiffFunction[U] = new DiffFunction[U] {
    override def calculate(u: U) = {
      val t = l.backward(u)
      val (obj,gu) = outer.calculate(t)
      (obj,l.forward(gu))
    }
  }
}

object DiffFunction {
  def withL2Regularization[T, I](d: DiffFunction[T],weight: Double)(implicit space: InnerProductModule[T, Double]) = new DiffFunction[T] {
    import space._
    override def gradientAt(x:T):T = {
      val grad = d.gradientAt(x)
      myGrad(grad,x)
    }

    override def valueAt(x:T) = {
      val v = d.valueAt(x)
      myValueAt(v, x)
    }

    private def myValueAt(v: Double, x:T) = {
      v + weight * (x dot x)/2
    }

    private def myGrad(g: T, x: T):T = {
      g + (x * weight)
    }

    override def calculate(x: T) = {
      val (v,grad) = d.calculate(x)
      (myValueAt(v, x), myGrad(grad,x))
    }
  }

  def withL2Regularization[T, I](d: BatchDiffFunction[T],weight: Double)(implicit space: InnerProductModule[T, Double]):BatchDiffFunction[T] = new BatchDiffFunction[T] {
    import space._
    override def gradientAt(x:T, batch: IndexedSeq[Int]):T = {
      val grad = d.gradientAt(x, batch)
      myGrad(grad,x)
    }

    override def valueAt(x:T, batch: IndexedSeq[Int]) = {
      val v = d.valueAt(x, batch)
      v + myValueAt(x)
    }

    private def myValueAt(x:T) = {
      weight * (x dot x)
    }

    private def myGrad(g: T, x: T) = {
      g + (x * weight)
    }

    override def calculate(x: T, batch: IndexedSeq[Int]) = {
      val (v,grad) = d.calculate(x, batch)
      (v + myValueAt(x), myGrad(grad,x))
    }

    def fullRange = d.fullRange
  }
}