bnmf-algs
|
Structure holding the results of EM procedures. More...
#include <online_EM_defs.hpp>
Public Member Functions | |
EMResult ()=default | |
Default constructor. More... | |
EMResult (matrix_t< T > S_pjk, matrix_t< T > S_ipk, matrix_t< T > X_full, matrix_t< T > logW, matrix_t< T > logH, vector_t< T > EM_bound) | |
Initialization constructor. More... | |
Public Attributes | |
matrix_t< T > | S_pjk |
Sum of the hidden tensor \(S\) along its first dimension, i.e. \(S_{+jk}\). More... | |
matrix_t< T > | S_ipk |
Sum of the hidden tensor \(S\) along its second dimension, i.e. \(S_{i+k}\). More... | |
matrix_t< T > | X_full |
Completed version of the incomplete matrix given as input to an EM algorithm. More... | |
matrix_t< double > | logW |
Matrix whose \((i, j)^{th}\) entry contains \(\log{W_{ij}}\). More... | |
matrix_t< double > | logH |
Matrix whose \((i, j)^{th}\) entry contains \(log{H_{ij}}\). More... | |
vector_t< double > | log_PS |
Vector containing EM bound computed after every iteration. More... | |
Structure holding the results of EM procedures.
T | Type of the matrix/tensor entries. |
|
default |
Default constructor.
Default constructor constructs every matrix/vector as empty.
|
inline |
Initialization constructor.
Every element is move initialized with the given value.
vector_t<double> bnmf_algs::bld::EMResult< T >::log_PS |
Vector containing EM bound computed after every iteration.
matrix_t<double> bnmf_algs::bld::EMResult< T >::logH |
Matrix whose \((i, j)^{th}\) entry contains \(log{H_{ij}}\).
matrix_t<double> bnmf_algs::bld::EMResult< T >::logW |
Matrix whose \((i, j)^{th}\) entry contains \(\log{W_{ij}}\).
matrix_t<T> bnmf_algs::bld::EMResult< T >::S_ipk |
Sum of the hidden tensor \(S\) along its second dimension, i.e. \(S_{i+k}\).
matrix_t<T> bnmf_algs::bld::EMResult< T >::S_pjk |
Sum of the hidden tensor \(S\) along its first dimension, i.e. \(S_{+jk}\).
matrix_t<T> bnmf_algs::bld::EMResult< T >::X_full |
Completed version of the incomplete matrix given as input to an EM algorithm.
EM algorithms take incomplete matrices (NaN entries are not known), and find the optimal values for those empty values. This is the matrix that contains the completed values for those entries.