Class InverseProblemOptions<T>

Namespace

AiDotNet.PhysicsInformed.Interfaces

Assembly

AiDotNet.dll

Configuration options for inverse problem PINN training.

public class InverseProblemOptions<T>

Type Parameters

T

The numeric type.

Inheritance

object

InverseProblemOptions<T>

Inherited Members

object.Equals(object)

object.Equals(object, object)

object.GetHashCode()

object.GetType()

object.MemberwiseClone()

object.ReferenceEquals(object, object)

object.ToString()

Properties

DataWeight

Weight for the observation data loss relative to physics loss.

public T? DataWeight { get; set; }

Property Value

T

Remarks

Higher weight: Trust observations more Lower weight: Trust physics more For noisy data, use lower weight

EstimateUncertainty

Whether to estimate parameter uncertainty.

public bool EstimateUncertainty { get; set; }

Property Value

bool

LogParameterHistory

Whether to log parameter estimates during training.

public bool LogParameterHistory { get; set; }

Property Value

bool

ParameterLearningRate

Learning rate for the unknown parameters (if separate rates are used).

public double ParameterLearningRate { get; set; }

Property Value

double

PriorMeans

Prior means for Bayesian regularization.

public T[]? PriorMeans { get; set; }

Property Value

T[]

PriorStandardDeviations

Prior standard deviations for Bayesian regularization.

public T[]? PriorStandardDeviations { get; set; }

Property Value

T[]

Regularization

The type of regularization to apply.

public InverseProblemRegularization Regularization { get; set; }

Property Value

InverseProblemRegularization

RegularizationStrength

Regularization strength (λ in the formulas above).

public T? RegularizationStrength { get; set; }

Property Value

T

Remarks

Too small: Solution may be unstable (overfitting to noise) Too large: Parameters biased toward prior/regularization Rule of thumb: Start with λ ≈ noise_level² / signal_level²

UncertaintySamples

Number of samples for uncertainty estimation (if enabled).

public int UncertaintySamples { get; set; }

Property Value

int

UseSeparateLearningRates

Whether to use separate learning rates for solution and parameters.

public bool UseSeparateLearningRates { get; set; }

Property Value

bool

Remarks

Parameters often need different learning rates than the neural network. Typically, parameters need smaller learning rates for stability.