Fixed bugs in dogleg and double_dogleg so that when the user
supplies an exact Hessian, it is used directly in the optimization
rather than being overridden by the BFGS approximation.
optimflex 0.1.6
Replaced the quadratic example with the Rosenbrock example.
Accordingly, removed dfp.R, which had failed to converge on this example
in certain environments.
optimflex 0.1.5
Bug Fixes
Fixed a convergence detection failure when the optimizer reaches the
exact solution in a single step. A post-step gradient check has been
added to all eight optimization algorithms.
Replaced test-rosenbrock.R with test-quadratic.R to verify
convergence on a simple quadratic function, consistent with the example
in the dfp documentation.
optimflex 0.1.4
Resubmission for CRAN.
Removed tests/testthat/test-rosenbrock.R, which was inadvertently
included in the previous submission and caused a convergence failure on
M1 Mac (ATLAS BLAS)
optimflex 0.1.3
Resubmission for CRAN.
Replaced the example in ‘dfp.R’ with a simpler one.
optimflex 0.1.1
Resubmission for CRAN.
Fixed CRAN check errors in multirel examples and
refined LaTeX documentation in dfp.R.
Implements a flexible suite of optimization algorithms (BFGS, DFP,
Dogleg, Newton, etc.).
Features strict convergence control with eight combinable
criteria.
Includes a built-in mechanism for final Hessian positive
definiteness verification.