libeblearn
ebl::gd_param Class Reference

#include <ebl_states.h>

Inheritance diagram for ebl::gd_param:
ebl::infer_param

List of all members.

Public Member Functions

 gd_param ()
 Empty constructor, set all variables to zero.
 gd_param (double leta, double ln, double l1, double l2, intg dtime, double iner, double a_v, intg a_p, double g_t)
 Constructor.

Public Attributes

double eta
 global step size
double n
 stopping criterion threshold
double decay_l1
 L1 regularizer coefficient.
double decay_l2
 L2 regularizer coefficient.
intg decay_time
 Time (in number of training samples) after which to decay values.
double inertia
 momentum term
double anneal_value
 annealing coefficient for the learning rate
intg anneal_period
 Number of training samples beetween two annealings.
double gradient_threshold
 threshold on square norm of gradient for stopping
int niter_done
 for debugging purpose

Detailed Description

A class that contains all the parameters for a stochastic gradient descent update, including step sizes, regularizer coefficients...


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