NonSmoothSolvers

Documentation for NonSmoothSolvers.

NonSmoothSolvers.OptimizationStateType
OptimizationState

Stores information after one iteration of the optimizer. Generic information is stored explicitly in the struct, custom information may be stored in the field additionalinfo::NamedTuple.

source
NonSmoothSolvers.optimize!Function
optimize!(state, pb, optimizer, optparams, tracestrategy, inittrace) -> Tuple{Any, DataStructures.Deque}

Call the optimizer on problem pb, with initial point initial_x. Returns a tuple containing the final iterate vector and a trace.

Features:

  • timing of the update_iterate method only;
  • saves basic information of each iteration in a vector of OptimizationState, the so-called trace;
  • the information saved at each iterate may be enriched by the user by providing a name and callback function via the optimstate_extension argument.

Example

getx(o, os) = os.x
optimstate_extensions = OrderedDict{Symbol, Function}(:x => getx)

optimize!(pb, o, xclose; optparams, optimstate_extensions)
source

Gradient sampling

<!– NonSmoothSolvers.update_iterate!(state, gs::GradientSampling, pb) –>

Non smooth BFGS

<!– NonSmoothSolvers.update_iterate! –>

<!– @autodocs --> <!-- Modules = [NonSmoothSolvers] --> <!-- pages = ["NonSmoothSolvers.jl", "optimize.jl", "solver_types.jl"] --> <!-- –>