.. _`topic-local_optimization_methods`: ========================== Local Optimization Methods ========================== Local optimization methods begin from an initial point, optionally specified by the user, and, broadly speaking, move "downhill" to the nearest local minimum. Dakota includes both gradient-based and derivative-free local optimizers. The :ref:`opt` section contains more information. Methods ------- Gradient-based ^^^^^^^^^^^^^^ - :dakkw:`method-optpp_cg` - :dakkw:`method-rol` - :dakkw:`method-dot_bfgs` - :dakkw:`method-dot_frcg` - :dakkw:`method-conmin_frcg` - :dakkw:`method-npsol_sqp` - :dakkw:`method-nlpql_sqp` - :dakkw:`method-dot_mmfd` - :dakkw:`method-dot_slp` - :dakkw:`method-dot_sqp` - :dakkw:`method-conmin_mfd` - :dakkw:`method-optpp_newton` - :dakkw:`method-optpp_q_newton` - :dakkw:`method-optpp_fd_newton` Derivative-free ^^^^^^^^^^^^^^^ - :dakkw:`method-optpp_pds` - :dakkw:`method-coliny_cobyla` - :dakkw:`method-coliny_pattern_search` - :dakkw:`method-coliny_solis_wets` - :dakkw:`method-asynch_pattern_search` - :dakkw:`method-surrogate_based_local` - :dakkw:`method-mesh_adaptive_search`