Package ffx.numerics.switching

Interface UnivariateSwitchingFunction

All Superinterfaces:
UnivariateDiffFunction
All Known Implementing Classes:
BellCurveSwitch, CompositeSwitch, ConstantSwitch, LinearDerivativeSwitch, MultiplicativeSwitch, PowerSwitch, SquaredTrigSwitch
public interface UnivariateSwitchingFunction extends UnivariateDiffFunction
A UnivariateSwitchingFunction describes a function of a single value (often lambda), where f(lb) = 0, f(ub) = 1, and df(x)/dx >= 0 for all x lb-ub.

Additionally, it's often useful for switching functions to have zero first and second derivatives at the lower and upper bound.

A number of methods exist to check for various properties of a switching function; these will often be implemented as simple return-boolean methods.

Author:
Jacob M. Litman, Michael J. Schnieders

  • Method Details

    • constantOutsideBounds

      boolean constantOutsideBounds()
      Remains 0 below the lower bound, and 1 above the upper bound (i.e. a multiplicative switch).
      Returns:
      df(x)/dx is zero outside range lb-ub.

    • getHighestOrderZeroDerivative

      int getHighestOrderZeroDerivative()
      The highest-order derivative that is zero at the bounds.
      Returns:
      Maximum order zero derivative at bounds.

    • getOneBound

      double getOneBound()
      Gets the one bound, where f(x) becomes one.
      Returns:
      One bound

    • getZeroBound

      double getZeroBound()
      Gets the zero bound, where f(x) becomes zero.
      Returns:
      Zero bound

    • highestOrderZeroDerivativeAtZeroBound

      default int highestOrderZeroDerivativeAtZeroBound()
      Returns the highest-order, guaranteed-zero derivative at the zero bound.
      Returns:
      Highest-order zero derivative at zero bound.

    • symmetricToUnity

      boolean symmetricToUnity()
      True if f(lb + delta) + f(ub - delta) = 1 for all delta between 0 and (ub - lb). For example, a power switch with beta 1 is symmetric to unity, as f(l) + f(1-l) = 1, but beta 2 produces a non-unity result, where f(0.5) + f(0.5) = 0.5.
      Returns:
      If symmetry produces unity result.

    • validOutsideBounds

      boolean validOutsideBounds()
      Remains in the range 0-1 outside the bounds. Implied to be true if constantOutsideBounds is true.
      Returns:
      min(f ( x)) = 0 and max(f(x)) = 1.