Previous  Up  Next

Module Lbfgs

module Lbfgs: sig .. end
The lbfgs library acts as a simple wrapper to the C LBFGS library written by Naoaki Okazaki.
type lbfgs_err =
| Success
| Stop
| AlreadyMinimized
| UnknownError
| LogicError
| OutOfMemory
| Canceled
| InvalidN
| InvalidNSSE
| InvalidXSSE
| InvalidEpsilon
| InvalidTestPeriod
| InvalidDelta
| InvalidLinesearch
| InvalidMinstep
| InvalidMaxstep
| InvalidFtol
| InvalidWolfe
| InvalidGtol
| InvalidXtol
| InvalidMaxLinesearch
| InvalidOrthantwise
| InvalidOrthantwiseStart
| InvalidOrthantwiseEnd
| OutOfInterval
| IncorrectTminmax
| RoundingError
| MinimumStep
| MaximumStep
| MaximumLinesearch
| MaximumIteration
| WidthTooSmall
| InvalidParameters
| IncreaseGradient
Type of error messages produced by LBFGS minimizer.
val errstring : lbfgs_err -> string
Convert an error type into a human-readable error string, mainly for debugging purposes.
type valgrad_func = float array -> float array -> float
Type of function callback used by the LBFGS minimizer. This function must computing the value and gradient of a function at a given point.
val minimize_l1 : float ->
       valgrad_func -> float array -> float -> int -> lbfgs_err * float
minimize_l1 c f x decrease maxiter finds the minimum of function f with L1 penalty weight c, starting its search at point x. Halts when the rate of decrease is less than decrease or after maxiter iterations. Returns an error code and minimum function value. x is modified to contain the minimizing point.
val minimize_l1_simple : float -> valgrad_func -> float array -> lbfgs_err * float
minimize_l1_simple c f x finds the minimum of function f with L1 penalty weight c, starting its search at point x. Uses default parameters for its convergence criteria. Returns an error code and minimum function value. x is modified to contain the minimizing point.
val minimize : valgrad_func -> float array -> float -> int -> lbfgs_err * float
minimize f x decrease maxiter finds the minimum of function f, starting its search at point x. Halts when minimum rate of decrease is less than decrease or after maxiter iterations. Returns an error code and minimum function value. x is modified to contain the minimizing point.
val minimize_simple : valgrad_func -> float array -> lbfgs_err * float
minimize_simple f x finds the minimum of function f, starting its search at point x. Uses default parameters for its convergence criteria. Returns an error code and minimum function value. x is modified to contain the minimizing point.