edu.stanford.nlp.optimization
Class QNMinimizer.Record
java.lang.Object
edu.stanford.nlp.optimization.QNMinimizer.Record
- Enclosing class:
- QNMinimizer
public class QNMinimizer.Record
- extends Object
The Record class is used to collect information about the function value
over a series of iterations. This information is used to determine
convergence, and to (attempt to) ensure numerical errors are not an issue.
It can also be used for plotting the results of the optimization routine.
- Author:
- akleeman
Method Summary |
void |
add(double val,
double[] grad,
double[] x,
int fevals)
|
double |
howLong()
Return the time in seconds since this class was created. |
void |
monitorX(double[] x)
|
void |
setEPS(double eps)
|
void |
setTOL(double tolerance)
|
void |
shutUp()
|
void |
start(double val,
double[] grad)
|
void |
start(double val,
double[] grad,
double[] x)
|
QNMinimizer.eState |
toContinue()
This function checks for convergence through first
order optimality, numerical convergence (i.e., zero numerical
gradient), and also by checking the average improvement. |
Methods inherited from class java.lang.Object |
clone, equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
QNMinimizer.Record
public QNMinimizer.Record()
QNMinimizer.Record
public QNMinimizer.Record(PrintWriter output)
QNMinimizer.Record
public QNMinimizer.Record(boolean beQuiet)
QNMinimizer.Record
public QNMinimizer.Record(boolean beQuiet,
Function monitor)
QNMinimizer.Record
public QNMinimizer.Record(boolean beQuiet,
Function monitor,
double tolerance)
QNMinimizer.Record
public QNMinimizer.Record(boolean beQuiet,
Function monitor,
double tolerance,
PrintWriter output)
QNMinimizer.Record
public QNMinimizer.Record(boolean beQuiet,
Function monitor,
double tolerance,
double eps)
setEPS
public void setEPS(double eps)
setTOL
public void setTOL(double tolerance)
start
public void start(double val,
double[] grad)
shutUp
public void shutUp()
start
public void start(double val,
double[] grad,
double[] x)
add
public void add(double val,
double[] grad,
double[] x,
int fevals)
monitorX
public void monitorX(double[] x)
toContinue
public QNMinimizer.eState toContinue()
- This function checks for convergence through first
order optimality, numerical convergence (i.e., zero numerical
gradient), and also by checking the average improvement.
- Returns:
- A value of the enumeration type
eState
which tells the
state of the optimization routine indicating whether the routine should
terminate, and if so why.
howLong
public double howLong()
- Return the time in seconds since this class was created.
- Returns:
- The time in seconds since this class was created.
Stanford NLP Group