v0.3

Added

• New ShiXieXuanNocedalForward.
• New ShiXieXuanNocedalGeneral.
• New benchmark_method.
• DumontetVignes: new get_ell_min_max.
• DumontetVignes, SteplemanWinarsky, GillMurraySaundersWright: new get_step_history.
• New SteplemanWinarkyInitialize: compute initial step for SteplemanWinarsky.

Changes

• DumontetVignes, SteplemanWinarsky, GillMurraySaundersWright: Renamed compute_step into find_step to avoid confusion.
• Removed SteplemanWinarsky.search_step_with_bisection: please use SteplemanWinarkyInitialize.find_initial_step().
• DerivativeBenchmark: new print and pretty-print.
• DerivativeBenchmark: create any problem from the test point and the interval. This can be useful to create a list of test points depending on each problem.
• ScaledExponentialProblem: change parametrization.
• DumontetVignes: parametrize depending on ell3 and ell4 instead of ell1 and ell2.
• GillMurraySaundersWright: Changed parametrization from relative precision
  of the function value into absolute precision to enable a zero function value.
• Set relative_precision to 1.0e-15 in all algorithms
• Set iteration_maximum to 53 in all algorithms
• PolynomialProblem: Fixed higher derivatives for alpha different from 2.
• DumontetVignes: Added a separate absolute precision parameter to manage cases where f(x) = 0.

Documentation

• DumontetVignes, SteplemanWinarsky, GillMurraySaundersWright: briefly introduce the method and its criteria.

Full Changelog: https://github.com/mbaudin47/numericalderivative/commits/v0.3

v0.2

Changes

• Renamed DerivativeBenchmark into DerivativeProblem.
  Other problems are renamed *Problem.
• Renamed GillMurraySaundersWrightExponentialDerivativeBenchmark
  into GMSWExponentialProblem.
• New benchmark problems: PolynomialProblem, InverseProblem,
  SXXNProblem1, SXXNProblem2, SXXNProblem3,
  SXXNProblem4, OliverProblem1, OliverProblem2, OliverProblem3.
• Removed FiniteDifferenceFormula and FiniteDifferenceOptimalStep.
  These features are equivalently provided by FirstDerivativeForward,
  FirstDerivativeCentral, SecondDerivativeCentral, ThirdDerivativeCentral.
• Renamed NumericalDerivative into FunctionWithArguments.
• Fixed Sphinx API help pages
• Added Examples sections into many docstrings.
• ThirdDerivativeCentral: Implement exact step and error for third derivative
• Scipy is a dependency
• SteplemanWinarsky.compute_step: separate hmin and hmax into two separate
  input arguments instead of a single list with 2 floats.
  This enables to set each parameter independently.
• New examples: use benchmark, finite difference formulas.

Added

• New GeneralFiniteDifference.

Full Changelog: https://github.com/mbaudin47/numericalderivative/commits/v0.2

v0.1

What's Changed

• New algorithms: DumontetVignes, SteplemanWinarsky, GillMurraySaundersWright
• New benchmark problems: DerivativeBenchmark, ExponentialDerivativeBenchmark,
  LogarithmicDerivativeBenchmark, SquareRootDerivativeBenchmark,
  AtanDerivativeBenchmark, SinDerivativeBenchmark,
  ScaledExponentialDerivativeBenchmark,
  GillMurraySaundersWrightExponentialDerivativeBenchmark.
• New benchmark features: BuildBenchmarkList
• New finite difference formulas and optimal steps:
  FiniteDifferenceFormula, FiniteDifferenceOptimalStep, NumericalDerivative
• New examples
• New Sphinx doc

Full Changelog: https://github.com/mbaudin47/numericalderivative/commits/v0.1