optimization

Type Members

1. case class CG_Rasmussen2(maxIters: Int = 256, minimumInterpolation: Double = 0.1, maximumEtrapolation: Double = 3.0, maximumEvaluationsPerLineSearch: Int = 20, ratio: Double = 10, sigma: Double = 0.1, rho: Double = 0.05) extends Minimizer with Product with Serializable

   Rasmussen's minimize.

2. case class CG_Rasmussen2_WithTermination(minimumInterpolation: Double = 0.1, maximumExtrapolation: Double = 3.0, maximumEvaluationsPerLineSearch: Int = 20, ratio: Double = 10, sigma: Double = 0.1, rho: Double = 0.05) extends Minimizer with Product with Serializable

   Rasmussen's minimize.

3. class CG_Rasmussen3 extends Minimizer

   Rasmussen's fmincg.

4. class ConjugateGradientDescent_HagerZhang extends Minimizer

   Implementation of the conjugate gradient descent as described in the article "A new conjugate gradient method wyth guaranteed descent and an efficient line search" by William W.

5. case class ConjugateGradientDescent_HagerZhangConfiguration(maxIters: Integer = scala.this.Predef.int2Integer(64), maxEvalsPerLineSearch: Integer = scala.this.Predef.int2Integer(16), delta: Double = 0.001, sigma: Double = 0.1, eta: Double = 0.01, epsilon: Double = 1.0E-9, theta: Double = 0.5, gamma: Double = 0.7) extends Product with Serializable

6. abstract class CubicInterpolationLineSearch extends NonlinearConjugateGradientDescent

   Line search algorithm as described in Carl Edward Rasmussen's unpublished (?) document "Function minimization using conjugate gradients: conj" (May 15 1996)

7. trait DifferentiableErrorFunctionFactory[X <: VectorSpace[X]] extends AnyRef

   Curried error function, which gets a target value and returns a differentiable function that represents some sort of error between the input and the target value.

8. trait DifferentiableFunction[V] extends (V) ⇒ Double

   Differentiable Double-valued function suitable for line-search algorithms.

9. class EarlyStopping[X, V] extends ResultSelector[X, V] with TerminationCriterion[X, Any]

   This termination criterion first just waits until numberOfInitialSteps have passed.

10. case class GradientDescent(maxIters: Int) extends Minimizer with Product with Serializable

11. class LimitNumberOfEvaluations extends TerminationCriterion[Any, (Int, Int)]

    Termination criterion that limits the number of line searches for CG-type optimization algorithms.

12. case class LimitNumberOfLineSearches(maxLineSearches: Int) extends TerminationCriterion[Any, (Int, Int)] with Product with Serializable

13. class LimitNumberOfSteps extends TerminationCriterion[Any, Int]

14. trait Minimizer extends AnyRef

15. abstract class NonlinearConjugateGradientDescent extends Minimizer

    Generic nonlinear conjugate gradient method for optimization of nonlinear twice differentiable functions which are bounded from below.

16. class NonlinearConjugateGradientDescent_Rasmussen extends CubicInterpolationLineSearch with PolakRibiere with SlopeRatioInitialStep

    Version of nonlinear conjugate gradient descent based on the minimize.

17. trait PolakRibiere extends NonlinearConjugateGradientDescent

18. class ResultSelector[X, F] extends AnyRef

    Result selector takes a look at a sequence of intermediate results of some optimization or search, and caches the best result it has seen so far.

19. trait SlopeRatioInitialStep extends NonlinearConjugateGradientDescent

20. trait TerminationCriterion[-IntermediateResult, -UsedResources] extends (IntermediateResult, UsedResources) ⇒ Boolean

    A fairly generic termination criterion that takes a look at the intermediate result achieved and resources used so far, and decides if an optimization procedure or some kind of search should be terminated.