Algorithms - Line Search

The OSTRICH line search optimization algorithm

Initial Publication:

Modified:

This group is used for configuration of the one-dimensional search algorithm (Brent or Golden-Section) that underlies an unconstrained numerical optimization procedure. Such algorithms are identified in the Overview above by the shaded entries in the “Line Search?” column. The general format (left side) and example (right side) of the Line Search group is given in Figure 1.

Begin1dSearch
1dSearchConvergeVal [cval]
1dSearchMethod Brent|GoldenSection  
End1dSearch
Begin1dSearch
1dSearchConvergeVal 0.0001
1dSearchMethod      GoldenSection  
End1dSearch

Figure 1: Format (left) and Example (right) of the Line Search Group

Where Begin1dSearch and Begin1dSearch are parsing tags that wrap the following set of algorithm configuration variables:

  • 1dSearchConvergeVal: Specifies the convergence value for the line search
  • 1dSearchMethod: Selects the line search algorithm.

The default convergence value is 0.0001 and the default search method is GoldenSection.