mic::neural_nets::loss::Loss< dtype > Class Template Reference

Abstract class representing a loss function. Defines interfaces. More...

#include <Loss.hpp>

Inheritance diagram for mic::neural_nets::loss::Loss< dtype >:

Public Member Functions
virtual dtype	calculateLoss (mic::types::MatrixPtr< dtype > target_y_, mic::types::MatrixPtr< dtype > predicted_y_)=0
	Function calculating loss - abstract. More...
virtual dtype	calculateMeanLoss (mic::types::MatrixPtr< dtype > target_y_, mic::types::MatrixPtr< dtype > predicted_y_)
	Calculates mean loss (i.e. divides the loss by the size of batch) - ACE for cross-entropy or MSE for regression. More...
virtual mic::types::MatrixPtr < dtype >	calculateGradient (mic::types::MatrixPtr< dtype > target_y_, mic::types::MatrixPtr< dtype > predicted_y_)=0
	Function calculating gradient - abstract. More...

Detailed Description

template<typename dtype = float>
class mic::neural_nets::loss::Loss< dtype >

Abstract class representing a loss function. Defines interfaces.

Author

tkornuta

Template Parameters

dtype

Template parameter denoting precision of variables.

Definition at line 41 of file Loss.hpp.

Member Function Documentation

template<typename dtype = float>

virtual mic::types::MatrixPtr<dtype> mic::neural_nets::loss::Loss< dtype >::calculateGradient ( mic::types::MatrixPtr< dtype >  target_y_, mic::types::MatrixPtr< dtype >  predicted_y_  )

pure virtual

Function calculating gradient - abstract.

Implemented in mic::neural_nets::loss::LogLikelihoodLoss< dtype >, mic::neural_nets::loss::CrossEntropyLoss< dtype >, mic::neural_nets::loss::SquaredErrorLoss< dtype >, mic::neural_nets::loss::SquaredErrorLoss< double >, and mic::neural_nets::loss::SquaredErrorLoss< float >.

template<typename dtype = float>

virtual dtype mic::neural_nets::loss::Loss< dtype >::calculateLoss ( mic::types::MatrixPtr< dtype >  target_y_, mic::types::MatrixPtr< dtype >  predicted_y_  )

pure virtual

Function calculating loss - abstract.

Implemented in mic::neural_nets::loss::LogLikelihoodLoss< dtype >, mic::neural_nets::loss::CrossEntropyLoss< dtype >, mic::neural_nets::loss::SquaredErrorLoss< dtype >, mic::neural_nets::loss::SquaredErrorLoss< double >, and mic::neural_nets::loss::SquaredErrorLoss< float >.

Referenced by mic::neural_nets::loss::Loss< float >::calculateMeanLoss().

template<typename dtype = float>

virtual dtype mic::neural_nets::loss::Loss< dtype >::calculateMeanLoss ( mic::types::MatrixPtr< dtype >  target_y_, mic::types::MatrixPtr< dtype >  predicted_y_  )

inlinevirtual

Calculates mean loss (i.e. divides the loss by the size of batch) - ACE for cross-entropy or MSE for regression.

Definition at line 51 of file Loss.hpp.

Referenced by TEST_F().

---

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

• src/loss/Loss.hpp