Class NelderMeadOptimizer<TInput, TMetric>

Namespace

SignalSharp.Optimization.NelderMead

Assembly

SignalSharp.dll

Implements the Nelder-Mead simplex optimization algorithm. It is a direct search method suitable for non-differentiable objective functions and does not use gradient information. This implementation handles parameter bounds by clamping.

public class NelderMeadOptimizer<TInput, TMetric> : IParameterOptimizer<TInput, TMetric> where TMetric : IFloatingPointIeee754<TMetric>

Type Parameters

TInput

The type of the input data provided to the objective function.

TMetric

The type of the objective metric (e.g., SSE), which must be a floating-point type.

Inheritance

object

NelderMeadOptimizer<TInput, TMetric>

Implements

IParameterOptimizer<TInput, TMetric>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Remarks

Instances keep mutable run state (evaluation counters) and are not thread-safe: use a separate instance per concurrent optimization.

Constructors

NelderMeadOptimizer(NelderMeadOptimizerOptions?, ILogger<NelderMeadOptimizer<TInput, TMetric>>?, Random?)

Implements the Nelder-Mead simplex optimization algorithm. It is a direct search method suitable for non-differentiable objective functions and does not use gradient information. This implementation handles parameter bounds by clamping.

public NelderMeadOptimizer(NelderMeadOptimizerOptions? options = null, ILogger<NelderMeadOptimizer<TInput, TMetric>>? logger = null, Random? random = null)

Parameters

options NelderMeadOptimizerOptions

logger ILogger<NelderMeadOptimizer<TInput, TMetric>>

random Random

Remarks

Instances keep mutable run state (evaluation counters) and are not thread-safe: use a separate instance per concurrent optimization.

Methods

Optimize(TInput, IEnumerable<ParameterDefinition>, Func<TInput, IReadOnlyDictionary<string, double>, ObjectiveEvaluation<TMetric>>)

Optimizes the objective function synchronously.

public OptimizationResult<TMetric> Optimize(TInput inputData, IEnumerable<ParameterDefinition> parametersToOptimize, Func<TInput, IReadOnlyDictionary<string, double>, ObjectiveEvaluation<TMetric>> objectiveFunction)

Parameters

inputData TInput

The input data passed to the objective function.

parametersToOptimize IEnumerable<ParameterDefinition>

The parameters to optimize, with unique names.

objectiveFunction Func<TInput, IReadOnlyDictionary<string, double>, ObjectiveEvaluation<TMetric>>

The objective function to minimize.

Returns

OptimizationResult<TMetric>

The optimization result.

Exceptions

ArgumentNullException

Thrown when the parameters or objective function are null.

ArgumentException

Thrown when parameter names are duplicated.

OptimizeAsync(TInput, IEnumerable<ParameterDefinition>, Func<TInput, IReadOnlyDictionary<string, double>, ObjectiveEvaluation<TMetric>>, CancellationToken)

Asynchronously optimizes the objective function. The computation is CPU-bound and executes synchronously on the calling thread; the returned task is already completed when the method returns. Cancellation is observed between objective evaluations.

public Task<OptimizationResult<TMetric>> OptimizeAsync(TInput inputData, IEnumerable<ParameterDefinition> parametersToOptimize, Func<TInput, IReadOnlyDictionary<string, double>, ObjectiveEvaluation<TMetric>> objectiveFunction, CancellationToken cancellationToken)

Parameters

inputData TInput

The input data passed to the objective function.

parametersToOptimize IEnumerable<ParameterDefinition>

The parameters to optimize, with unique names.

objectiveFunction Func<TInput, IReadOnlyDictionary<string, double>, ObjectiveEvaluation<TMetric>>

The objective function to minimize.

cancellationToken CancellationToken

Token to cancel the optimization process.

Returns

Task<OptimizationResult<TMetric>>

A task containing the optimization result.

Exceptions

ArgumentNullException

Thrown when the parameters or objective function are null.

ArgumentException

Thrown when parameter names are duplicated.