txeo::Loss< T > Class Template Reference

Computes error metrics between predicted and validation tensors. More...

#include <Loss.h>

Collaboration diagram for txeo::Loss< T >:

Public Member Functions
	Loss (const Loss &)=default
	Loss (Loss &&)=default
Loss &	operator= (const Loss &)=default
Loss &	operator= (Loss &&)=default
	~Loss ()=default
	Loss (txeo::Tensor< T > &&label, txeo::LossFunc func=txeo::LossFunc::MSE)
	Construct a new Loss object from a label rvalue.
	Loss (const txeo::Tensor< T > &label, txeo::LossFunc func=txeo::LossFunc::MSE)
	Construct a new Loss object.
T	get_loss (const txeo::Tensor< T > &pred) const
	Compute loss using currently selected function.
void	set_loss (txeo::LossFunc func)
	Set the active loss function.
T	mean_squared_error (const txeo::Tensor< T > &pred) const
	Compute Mean Squared Error (MSE)
T	mean_absolute_error (const txeo::Tensor< T > &pred) const
	Compute Mean Absolute Error (MAE)
T	mean_squared_logarithmic_error (const txeo::Tensor< T > &pred) const
	Compute Mean Squared Logarithmic Error (MSLE)
T	log_cosh_error (const txeo::Tensor< T > &pred) const
	Compute Log-Cosh Error (LCHE)
const txeo::Tensor< T > &	label () const
Shorthand Aliases
T	mse (const txeo::Tensor< T > &pred) const
T	mae (const txeo::Tensor< T > &pred) const
T	msle (const txeo::Tensor< T > &pred) const
T	lche (const txeo::Tensor< T > &pred) const

Detailed Description

template<typename T>
class txeo::Loss< T >

Computes error metrics between predicted and validation tensors.

Supports multiple loss functions that can be selected at runtime:

• MSE (Mean Squared Error)
• MAE (Mean Absolute Error)
• MSLE (Mean Squared Logarithmic Error)
• LCHE (Log-Cosh Error)

Template Parameters

T	Numeric type of tensor elements (float/double recommended)

Note

All operations validate tensor shape compatibility between predictions and validation data;

First axis's dimension of the tensors involved must refer to number of samples

Definition at line 30 of file Loss.h.

Constructor & Destructor Documentation

Loss() [1/4]

template<typename T >

txeo::Loss< T >::Loss ( const Loss< T > &  )

default

Loss() [2/4]

template<typename T >

txeo::Loss< T >::Loss ( Loss< T > &&  )

default

~Loss()

template<typename T >

txeo::Loss< T >::~Loss ( )

default

Loss() [3/4]

template<typename T >

txeo::Loss< T >::Loss ( txeo::Tensor< T > &&  label, txeo::LossFunc  func = txeo::LossFunc::MSE  )

explicit

Construct a new Loss object from a label rvalue.

Parameters

label	Labe tensor containing ground truth values
func	Initial loss function (default: MSE)

Example:

// Create with MAE as default loss
txeo::Loss<float> loss(validation_tensor, txeo::LossFunc::MAE);

Loss() [4/4]

template<typename T >

txeo::Loss< T >::Loss ( const txeo::Tensor< T > &  label, txeo::LossFunc  func = txeo::LossFunc::MSE  )

inlineexplicit

Construct a new Loss object.

Parameters

label	Label tensor containing ground truth values
func	Initial loss function (default: MSE)

Example:

// Create with MAE as default loss
txeo::Loss<float> loss(validation_tensor, txeo::LossFunc::MAE);

Definition at line 64 of file Loss.h.

        : Loss{label.clone(), func} {};

Member Function Documentation

get_loss()

template<typename T >

T txeo::Loss< T >::get_loss

(

const txeo::Tensor< T > &

pred

)

const

Compute loss using currently selected function.

Parameters

pred	Prediction tensor

Returns

T Calculated loss value

Exceptions

LossError

If shapes mismatch or invalid input values

Example:

auto error = loss.get_loss(predictions);
std::cout << "Current loss: " << error << std::endl;

label()

template<typename T >

const txeo::Tensor< T > & txeo::Loss< T >::label ( ) const

inline

Definition at line 136 of file Loss.h.

{ return _label; }

lche()

template<typename T >

T txeo::Loss< T >::lche ( const txeo::Tensor< T > &  pred ) const

inline

See also

mean_squared_logarithmic_error

Definition at line 152 of file Loss.h.

                                            {
      return log_cosh_error(pred);
    }; 

Here is the call graph for this function:

log_cosh_error()

template<typename T >

T txeo::Loss< T >::log_cosh_error

(

const txeo::Tensor< T > &

pred

)

const

Compute Log-Cosh Error (LCHE)

\[ LCHE = \frac{1}{N}\sum_{i=1}^{N}\log(\cosh(y_i - \hat{y}_i)) \]

Parameters

pred	Prediction tensor

Returns

T LCHE value

Here is the caller graph for this function:

mae()

template<typename T >

T txeo::Loss< T >::mae ( const txeo::Tensor< T > &  pred ) const

inline

See also

mean_squared_error

Definition at line 144 of file Loss.h.

                                           {
      return mean_absolute_error(pred);
    }; 

Here is the call graph for this function:

mean_absolute_error()

template<typename T >

T txeo::Loss< T >::mean_absolute_error

(

const txeo::Tensor< T > &

pred

)

const

Compute Mean Absolute Error (MAE)

\[ MAE = \frac{1}{N}\sum_{i=1}^{N}|y_i - \hat{y}_i| \]

Parameters

pred	Prediction tensor

Returns

T MAE value

Here is the caller graph for this function:

mean_squared_error()

template<typename T >

T txeo::Loss< T >::mean_squared_error

(

const txeo::Tensor< T > &

pred

)

const

Compute Mean Squared Error (MSE)

\[ MSE = \frac{1}{N}\sum_{i=1}^{N}(y_i - \hat{y}_i)^2 \]

Parameters

pred	Prediction tensor

Returns

T MSE value

Here is the caller graph for this function:

mean_squared_logarithmic_error()

template<typename T >

T txeo::Loss< T >::mean_squared_logarithmic_error

(

const txeo::Tensor< T > &

pred

)

const

Compute Mean Squared Logarithmic Error (MSLE)

\[ MSLE = \frac{1}{N}\sum_{i=1}^{N}(\log(1+y_i) - \log(1+\hat{y}_i))^2 \]

Parameters

pred	Prediction tensor

Returns

T MSLE value

Exceptions

LossError

If any values are negative

Here is the caller graph for this function:

mse()

template<typename T >

T txeo::Loss< T >::mse ( const txeo::Tensor< T > &  pred ) const

inline

Definition at line 140 of file Loss.h.

                                           {
      return mean_squared_error(pred);
    }; 

Here is the call graph for this function:

msle()

template<typename T >

T txeo::Loss< T >::msle ( const txeo::Tensor< T > &  pred ) const

inline

See also

mean_absolute_error

Definition at line 148 of file Loss.h.

                                            {
      return mean_squared_logarithmic_error(pred);
    }; 

Here is the call graph for this function:

operator=() [1/2]

template<typename T >

Loss & txeo::Loss< T >::operator= ( const Loss< T > &  )

default

operator=() [2/2]

template<typename T >

Loss & txeo::Loss< T >::operator= ( Loss< T > &&  )

default

set_loss()

template<typename T >

void txeo::Loss< T >::set_loss

(

txeo::LossFunc

func

)

Set the active loss function.

Parameters

func	Loss function to use for subsequent calculations

Example:

// Switch to log-cosh error
loss.set_loss(txeo::LossFunc::LCHE);

---

The documentation for this class was generated from the following file:

• include/txeo/Loss.h