DifferentiableStateSpaceModels

Thread-local caches, automatically generating new caches when required.

ThreadLocalCache(cache)

Examples

julia> caches = ThreadLocalCache(rand(5))

julia> caches()  # get a local cache for the thread 

Helper to allocate threadsafe cache objects consistent with any AbstractPerturbationModel for use with generate_perturbation.

Initializes with a cache for the current thread.

allocate_cache(m)

Examples

julia> cache = allocate_cache(m);

julia> sol = generate_perturbation(m; cache)  # cache can be used in different threads

Solve a model, optionally reusing a threadsafe cache.

generate_perturbation(m, p; p_f, cache, settings)

Details

Returns the solution

Get a threadsafe cache instance, and determine if the current p and p_f match it.

NOTE: in the future there could be memoization of multiple caches, but right now there is one per thread.

get_threadsafe_cache(cache, m, p, p_f)
get_threadsafe_cache(cache, m, p, p_f, settings)

Examples

julia> cache = allocate_cache(m);

julia> c, hash_match_all, hash_match_ss, hash_match_perturbation = get_threadsafe_cache(cache, m, p, p_f)

Return model from a stored module, creating as necessary . If the module is

• already included, it just returns the Module
• saved to disk but not already included, it does include
• not saved to disk, it saves it, includes it

The specification of a model_generator has a sepcification like H, mod_vals = model_generator() where H is an MTK expression the mod_vals can be splatted into the FirstOrderPerturbationModel or SecondOrderPerturbationModel constructor.

The name of the stored module is the name of model_generator function and the order.

Examples

julia> m = @include_example_module(Examples.rbc_observables_benchmark)  # stores or loads `.function_cache/rbc_observables_1.jl` 

julia> m = @include_example_module(Examples.rbc_observables_benchmark, 2) # second order, loads `.function_cache/rbc_observables_2.jl`