Template Class Model

Defined in File model.hpp

Nested Relationships

Nested Types

Inheritance Relationships

Base Type

public otyca::ModelBase (Class ModelBase)

Derived Type

public otyca::CompositeModel< N > (Template Class CompositeModel)

Class Documentation

template<std::size_t N> class Model : public virtual otyca::ModelBase 

Subclassed by otyca::CompositeModel< N >

Public Functions

virtual ~Model() = default

inline virtual double evaluate(const std::array<double, N> &x) const

virtual std::vector<double> evaluate(const std::vector<std::array<double, N>> &x) const = 0

template<std::size_t M = N> inline std::enable_if_t<M == 1, double> evaluate(double x) const

template<std::size_t M = N> inline std::enable_if_t<M == 1, std::vector<double>> evaluate(const std::vector<double> &x) const

inline virtual void ukf(const CalibrationInput &ci)

inline virtual double chi2(const CalibrationInput &ci) const

inline virtual double bias(const CalibrationInput &ci) const

inline virtual std::vector<size_t> violations(const CalibrationInput &ci) const

inline virtual std::vector<double> get_regularizer_penalties() const

inline virtual void calibrate(const CalibrationInput &ci, const CalibrationConfig &cc = CalibrationConfig())

inline bool is_calibrated() const

inline std::optional<CalibrationStatus> get_last_calibration_status() const

inline double get_last_calibration_reduced_chi2() const

inline double get_last_calibration_aic() const

inline double get_last_calibration_bic() const

inline int get_last_calibration_iterations() const

inline int get_last_calibration_residuals() const

inline void reset_calibration_status()

struct CalibrationConfig

Public Members

int maxfev = 10000: Maximum number of function evaluations.

double ftol = 1e-6: Tolerance for relative error in the sum of squares.

double xtol = 1e-6: Tolerance for relative error in parameter values.

double gtol = 1e-6: Tolerance for orthogonality of residuals and Jacobian.

double epsfcn = 1e-6: Step size used for finite-difference Jacobian approximation.

double factor = 1000.0: Initial step bound for the Levenberg-Marquardt algorithm.

struct CalibrationInput

Represents a calibration dataset with multi-dimensional input features, target values, and uncertainties.

This struct stores multi-dimensional input vectors (represented as arrays of size N), target output values (y), and uncertainties (dy) associated with the output targets. The struct includes a method for checking consistency between the sizes of the input, output, and uncertainty vectors.

Public Functions

inline bool size_consistent() const

Checks whether the sizes of the input, output, and uncertainty vectors are consistent.

Ensures that the number of input samples matches the number of target values and their uncertainties

Returns:: True if all vectors have consistent sizes; false otherwise.

inline void clear()

inline void serialize(const std::string &filename) const

Serializes the contents of the CalibrationInput object to a binary file.

The function writes the size of the data followed by the raw bytes of the x, y, and dy vectors to the specified file using binary format.

Parameters:: filename – Path to the output file.
Throws:: std::runtime_error – if the file cannot be opened.

inline void deserialize(const std::string &filename)

Deserializes the contents of a CalibrationInput object from a binary file.

The function reads the binary-encoded data from the specified file and populates the x, y, and dy vectors.

Parameters:: filename – Path to the input file.
Throws:: std::runtime_error – if the file cannot be opened or if the read fails.

inline CalibrationInput subset(int sample_interval) const

Returns a sampled subset of the calibration data.

This function flattens the calibration data stored in x, y, and dy, applies uniform downsampling based on the provided sample_interval, and then reconstructs a new CalibrationInput object with the sampled data.

Parameters:: sample_interval – Sampling stride: every sample_interval-th data point will be retained. Must be ≥ 1.
Returns:: A new CalibrationInput instance containing the sampled data.

Public Members

std::vector<std::array<double, N>> x: Multi-dimensional input values. Each entry in the vector is an array of size N representing a set of input features for a sample.

std::vector<double> y: Target output values. A vector of target values associated with each input sample.

std::vector<double> dy: Uncertainties associated with the target values. A vector of uncertainties (standard deviations) corresponding to the target values.

struct CalibrationStatus

Public Functions

inline double reduced_chi2() const

inline double aic() const

inline double bic() const

inline std::string message()

Returns a human-readable explanation of why the optimization stopped.

This function decodes the info value and provides a descriptive reason for the termination of the optimization process.

Returns:: A string describing the reason for stopping.

inline bool converged() const

Public Members

double chi2 = NAN: sum of differences between observed and expected, each divided by the standard deviation of observed and squared

double cost2 = NAN: L2 norm of cost vector evaluated at fitted parameters.

int residuals = 0: Number of residuals.

int iterations = 0: Number of iterations.

int parameters = 0: Number of model parameters.

int info = 0: See stop_reason.