SimulatedAnnealing (bristlecone) | bristlecone

bristlecone

SimulatedAnnealing Module

Namespace: Bristlecone.Optimisation
Assembly: Bristlecone.dll
Parent Module: MonteCarlo

A meta-heuristic that approximates a global optimium by simulating slow cooling as a slow decrease in the probability of temporarily accepting worse solutions.

Types and nested modules

Type/Module	Description
CoolingSchemes	Cooling schemes dictate the conditions under which the temperature is cooled during simulated annealing.
Machines
Tuning
AnnealSettings	Represents configurable settings of an annealing procedure that supports (a) heating, followed by (b) annealing.

Functions and values

Function or value	Description
`anneal writeOut saEnd tempFloor cool markov temperature (arg7, arg7) previousBests iteration` Full Usage: `anneal writeOut saEnd tempFloor cool markov temperature (arg7, arg7) previousBests iteration` Parameters: writeOut : `LogEvent -> unit` saEnd : `EndCondition` tempFloor : `float option` cool : `float -> int -> float` markov : `float -> Solution -> Solution list` temperature : `float` arg6 : `float<MeasureProduct<-logL, MeasureOne>>` arg7 : `Point` previousBests : `OptimisationTrace list` iteration : `int<MeasureProduct<iteration, MeasureOne>>` Returns: `OptimisationTrace list`	Cool between homoegenous markov chains according to `cool` schedule. Each anneal recursion begins from the end of the previous markov chain. writeOut : `LogEvent -> unit` saEnd : `EndCondition` tempFloor : `float option` cool : `float -> int -> float` markov : `float -> Solution -> Solution list` temperature : `float` arg6 : `float<MeasureProduct<-logL, MeasureOne>>` arg7 : `Point` previousBests : `OptimisationTrace list` iteration : `int<MeasureProduct<iteration, MeasureOne>>` Returns: `OptimisationTrace list`
`classicalSimulatedAnnealing scale tDependentProposal settings` Full Usage: `classicalSimulatedAnnealing scale tDependentProposal settings` Parameters: scale : `float<MeasureProduct<optim-space, MeasureOne>>` tDependentProposal : `bool` settings : `AnnealSettings` Returns: `Optimiser`	Candidate distribution: Gaussian univariate [] Probability: Boltzmann Machine scale : `float<MeasureProduct<optim-space, MeasureOne>>` tDependentProposal : `bool` settings : `AnnealSettings` Returns: `Optimiser`
`fastSimulatedAnnealing scale tDependentProposal settings` Full Usage: `fastSimulatedAnnealing scale tDependentProposal settings` Parameters: scale : `float<MeasureProduct<optim-space, MeasureOne>>` tDependentProposal : `bool` settings : `AnnealSettings` Returns: `Optimiser`	Candidate distribution: Cauchy univariate [] Probability: Bottzmann Machine scale : `float<MeasureProduct<optim-space, MeasureOne>>` tDependentProposal : `bool` settings : `AnnealSettings` Returns: `Optimiser`
`heat write ceiling endAcceptanceRate heatingSchedule markov (arg6, arg6) history temperature` Full Usage: `heat write ceiling endAcceptanceRate heatingSchedule markov (arg6, arg6) history temperature` Parameters: write : `LogEvent -> unit` ceiling : `float option` endAcceptanceRate : `float` heatingSchedule : `float -> float` markov : `float -> Solution -> (float<MeasureProduct<-logL, MeasureOne>> * Point) list` arg5 : `float<MeasureProduct<-logL, MeasureOne>>` arg6 : `Point` history : `OptimisationTrace list` temperature : `float` Returns: `OptimisationTrace list * float`	Heat up temperature until acceptance rate of bad moves is above the threshold `endAcceptanceRate`. If it becomes impossible to propose a move during heating, then heating ends. write : `LogEvent -> unit` ceiling : `float option` endAcceptanceRate : `float` heatingSchedule : `float -> float` markov : `float -> Solution -> (float<MeasureProduct<-logL, MeasureOne>> * Point) list` arg5 : `float<MeasureProduct<-logL, MeasureOne>>` arg6 : `Point` history : `OptimisationTrace list` temperature : `float` Returns: `OptimisationTrace list * float`
`markovChain writeOut atEnd propose probability random f temperature (arg8, arg8)` Full Usage: `markovChain writeOut atEnd propose probability random f temperature (arg8, arg8)` Parameters: writeOut : `LogEvent -> unit` atEnd : `Solution list -> int<MeasureProduct<iteration, MeasureOne>> -> OptimStopReason` propose : `Point -> Point` probability : `'a -> float<MeasureProduct<-logL, MeasureOne>> -> float<MeasureOne>` random : `Random` f : `Objective` temperature : `'a` arg7 : `float<MeasureProduct<-logL, MeasureOne>>` arg8 : `Point` Returns: `Solution list`	Run a homogenous Markov chain recursively until an end condition - `atEnd` - is met. writeOut : `LogEvent -> unit` atEnd : `Solution list -> int<MeasureProduct<iteration, MeasureOne>> -> OptimStopReason` propose : `Point -> Point` probability : `'a -> float<MeasureProduct<-logL, MeasureOne>> -> float<MeasureOne>` random : `Random` f : `Objective` temperature : `'a` arg7 : `float<MeasureProduct<-logL, MeasureOne>>` arg8 : `Point` Returns: `Solution list`
`simulatedAnnealing initialScale settings endCondition machine jump cool random writeOut domain f` Full Usage: `simulatedAnnealing initialScale settings endCondition machine jump cool random writeOut domain f` Parameters: initialScale : `float<MeasureProduct<optim-space, MeasureOne>>[]` settings : `AnnealSettings` endCondition : `EndCondition` machine : `float -> float<MeasureProduct<-logL, MeasureOne>> -> float` jump : `Random -> float<MeasureProduct<optim-space, MeasureOne>> -> float -> unit -> float<MeasureProduct<optim-space, MeasureOne>>` cool : `float -> float -> int -> float` random : `Random` writeOut : `LogEvent -> unit` domain : `Domain` f : `TypedTensor<Vector, MeasureProduct<optim-space, MeasureOne>> -> TypedTensor<Scalar, MeasureProduct<-logL, MeasureOne>>` Returns: `OptimisationTrace list * float<MeasureProduct<optim-space, MeasureOne>>[]`	initialScale : `float<MeasureProduct<optim-space, MeasureOne>>[]` settings : `AnnealSettings` endCondition : `EndCondition` machine : `float -> float<MeasureProduct<-logL, MeasureOne>> -> float` jump : `Random -> float<MeasureProduct<optim-space, MeasureOne>> -> float -> unit -> float<MeasureProduct<optim-space, MeasureOne>>` cool : `float -> float -> int -> float` random : `Random` writeOut : `LogEvent -> unit` domain : `Domain` f : `TypedTensor<Vector, MeasureProduct<optim-space, MeasureOne>> -> TypedTensor<Scalar, MeasureProduct<-logL, MeasureOne>>` Returns: `OptimisationTrace list * float<MeasureProduct<optim-space, MeasureOne>>[]`
`tryMove propose probability random f tries (l1, theta1)` Full Usage: `tryMove propose probability random f tries (l1, theta1)` Parameters: propose : `Point -> Point` probability : `float<MeasureProduct<-logL, MeasureOne>> -> float<MeasureOne>` random : `Random` f : `Objective` tries : `int` l1 : `float<MeasureProduct<-logL, MeasureOne>>` theta1 : `Point` Returns: `Solution option`	Jump based on a proposal function and probability function propose : `Point -> Point` probability : `float<MeasureProduct<-logL, MeasureOne>> -> float<MeasureOne>` random : `Random` f : `Objective` tries : `int` l1 : `float<MeasureProduct<-logL, MeasureOne>>` theta1 : `Point` Returns: `Solution option`