LBFGSLineSearchPrms
The LBFGSLineSearchPrms
case class is used to group together all parameters that control the line search routine used by the L-BFGS optimization process.
Value parameters
- defaultStep
-
The default step selected as the initial for the line search routine. The default value is 1.0.
- ftol
-
Controls the accuracy of the line search routine. Should be greater than zero and smaller than 0.5. The default value is 1e-4.
- gtol
-
Controls the accuracy of the line search routine. If the function and gradient evaluations are inexpensive with respect to the cost of the iteration (which is sometimes the case when solving very large problems), it may be advantageous to set this parameter to a small value (e.g: 0.1). This parameter should be greater than the
ftol
parameter (default value of 1e-4) and smaller than 1.0. The default value is 0.9. - maxLineSearch
-
Maximum number of trials for the line search. Controls the number of function and gradient evaluations per iteration for the line search routine. The default value is 40.
- maxStep
-
Maximum step of the line search routine. Does not need to be modified unless the exponents are too large for the machine being used, or unless the problem is extremely badly scaled (in which case the exponents should be increased). The default value is 1e20.
- minStep
-
Minimum step of the line search routine. Does not need to be modified unless the exponents are too large for the machine being used, or unless the problem is extremely badly scaled (in which case the exponents should be increased). The default value is 1e-20.
- wolfe
-
A coefficient for the Wolfe condition. Only used when a backtracking line-search algorithm that relies on the Wolfe condition is chosen for the
LBFGSLineSearchAlg
lineSearch
param (e.g:LBFGSLineSearchAlg.BacktrackingWolfe
orLBFGSLineSearchAlg.BacktrackingStrongWolfe
). Should be greater than theftol
parameter and smaller than 1.0. The default value is 0.9. - xtol
-
The machine precision for floating-point values. Must be a positive value set by a client program to estimate the machine precision. The L-BFGS optimization will terminate with the return code
LBFGSReturnCode.RoundingError
if the relative width of the interval of uncertainty is less than this parameter. The default value is 1.0e-16.
Attributes
- Graph
-
- Supertypes
-
trait Serializabletrait Producttrait Equalsclass Objecttrait Matchableclass Any