hex
Class DLSM.GeneralizedGradientSolver
java.lang.Object
water.Iced
hex.DLSM.LSMSolver
hex.DLSM.GeneralizedGradientSolver
- All Implemented Interfaces:
- java.lang.Cloneable, Freezable
- Enclosing class:
- DLSM
public static final class DLSM.GeneralizedGradientSolver
- extends DLSM.LSMSolver
Generalized gradient solver for solving LSM problem with combination of L1 and L2 penalty.
Method Summary |
java.lang.String |
name()
|
boolean |
solve(hex.DGLM.Gram gram,
double[] newBeta)
|
com.google.gson.JsonObject |
toJson()
|
Methods inherited from class java.lang.Object |
equals, finalize, getClass, hashCode, notify, notifyAll, toString, wait, wait, wait |
_kappa
public final double _kappa
_betaEps
public final double _betaEps
MAX_ITER
public static final int MAX_ITER
- See Also:
- Constant Field Values
EPS
public static final double EPS
- See Also:
- Constant Field Values
DLSM.GeneralizedGradientSolver
public DLSM.GeneralizedGradientSolver(double lambda,
double alpha)
DLSM.GeneralizedGradientSolver
public DLSM.GeneralizedGradientSolver(double lambda,
double alpha,
double betaEps)
solve
public boolean solve(hex.DGLM.Gram gram,
double[] newBeta)
- Specified by:
solve
in class DLSM.LSMSolver
- Parameters:
xx:
- gram matrix. gaussian: X'X, binomial:(1/4)X'Xxy:
- -X'y (LSM) l or -(1/4)X'(XB + (y-p)/(p*1-p))(IRLSMyy:
- 0.5*y'*y gaussian, 0.25*z'*z IRLSMbeta:
- previous vector of coefficients, will be modified/destroyednewBeta:
- resulting vector of coefficients
- Returns:
- true if converged
toJson
public com.google.gson.JsonObject toJson()
- Specified by:
toJson
in class DLSM.LSMSolver
name
public java.lang.String name()
- Specified by:
name
in class DLSM.LSMSolver