Initialization
Before we can use update! we have to initialize with SimpleSolvers.initialize![1].
Similar to update!, SimpleSolvers.initialize! returns the first input argument as output. Examples include:
SimpleSolvers.initialize!(::SimpleSolvers.NonlinearSolverCache, ::AbstractVector),SimpleSolvers.initialize!(::SimpleSolvers.NewtonOptimizerCache, ::AbstractVector).
We demonstrate this for an instance of SimpleSolvers.NewtonOptimizerCache[2]:
x = rand(3)
cache = SimpleSolvers.NewtonOptimizerCache(x)
initialize!(cache, x)
cache.g3-element Vector{Float64}:
NaN
NaN
NaNReasoning behind Initialization with NaNs
We initialize with NaNs instead of with zeros (or other values) as this clearly divides the initialization from the numerical operations (which are done with update!).
Alloc Functions
- 1The different methods for
SimpleSolvers.initialize!are however often called with the constructor of astruct(e.g. forSimpleSolvers.NewtonOptimizerCache). - 2Here we remark that
SimpleSolvers.NewtonOptimizerCachehas five keys:x,x̄,δ,gandrhs. All of them are initialized withNaNs except forx. We also remark that the constructor, which is called by providing a single vectorxas input argument, internally also callsSimpleSolvers.initialize!.