LIBIRWLS
|
Functions to train a full SVM using the IRWLS algorithm. More...
#include "IOStructures.h"
Go to the source code of this file.
Functions | |
int * | rpermute (int n) |
Random permutation of n elements. More... | |
double * | subIRWLS (svm_dataset dataset, properties props, double *GIN, double *e, double *beta) |
IRWLS procedure on a Working Set. More... | |
double * | trainFULL (svm_dataset dataset, properties props) |
It trains a full SVM with a training set. More... | |
void | printPIRWLSInstructions (void) |
Print Instructions. More... | |
properties | parseTrainPIRWLSParameters (int *argc, char ***argv) |
It parses the command line. More... | |
model | calculatePIRWLSModel (properties props, svm_dataset dataset, double *beta) |
It converts the result into a model struct. More... | |
Functions to train a full SVM using the IRWLS algorithm.
For a detailed description of the algorithm and its parameters read the following paper:
Pérez-Cruz, F., Alarcón-Diana, P. L., Navia-Vázquez, A., & Artés-Rodríguez, A. (2001). Fast Training of Support Vector Classifiers. In Advances in Neural Information Processing Systems (pp. 734-740)
For a detailed description about the parallelization read the following paper:
Díaz-Morales, R., & Navia-Vázquez, Á. (2016). Efficient parallel implementation of kernel methods. Neurocomputing, 191, 175-186.
model calculatePIRWLSModel | ( | properties | props, |
svm_dataset | dataset, | ||
double * | beta | ||
) |
It converts the result into a model struct.
After the training of a SVM using the IRWLS procedure, this function build a struct with the information and returns it.
props | The training parameters. |
dataset | The training set. |
beta | The weights of the classifier. |
properties parseTrainPIRWLSParameters | ( | int * | argc, |
char *** | argv | ||
) |
It parses the command line.
It parses input command line to extract the parameters.
argc | The number of words of the command line. |
argv | The list of words of the command line. |
void printPIRWLSInstructions | ( | void | ) |
Print Instructions.
It shows PIRWLS-train command line instructions in the standard output.
int* rpermute | ( | int | n | ) |
Random permutation of n elements.
It crates a random permutation of n elements.
n | The number of elementos in the permutation. |
double* subIRWLS | ( | svm_dataset | dataset, |
properties | props, | ||
double * | GIN, | ||
double * | e, | ||
double * | beta | ||
) |
IRWLS procedure on a Working Set.
For a detailed description of the algorithm and its parameters read the following paper:
Pérez-Cruz, F., Alarcón-Diana, P. L., Navia-Vázquez, A., & Artés-Rodríguez, A. (2001). Fast Training of Support Vector Classifiers. In Advances in Neural Information Processing Systems (pp. 734-740)
For a detailed description about the parallelization read the following paper:
Díaz-Morales, R., & Navia-Vázquez, Á. (2016). Efficient parallel implementation of kernel methods. Neurocomputing, 191, 175-186.
dataset | The training dataset. |
props | The strut of training properties. |
GIN | The classification effect of the inactive set. |
e | The current error on every training data. |
beta | The bias term of the classification function. |
double* trainFULL | ( | svm_dataset | dataset, |
properties | props | ||
) |
It trains a full SVM with a training set.
For a detailed description of the algorithm and its parameters read the following paper:
Pérez-Cruz, F., Alarcón-Diana, P. L., Navia-Vázquez, A., & Artés-Rodríguez, A. (2001). Fast Training of Support Vector Classifiers. In Advances in Neural Information Processing Systems (pp. 734-740)
For a detailed description about the parallelization read the following paper:
Díaz-Morales, R., & Navia-Vázquez, Á. (2016). Efficient parallel implementation of kernel methods. Neurocomputing, 191, 175-186.
It trains a full SVM using a training set and the training parameters.
dataset | The training set. |
props | The values of the training parameters. |