class
TruncatedNewtonMinimizer[T, H] extends Minimizer[T, SecondOrderFunction[T, H]] with Logging
Instance Constructors
-
new
TruncatedNewtonMinimizer(maxIterations: Int = -1, tolerance: Double = 1.0E-6, l2Regularization: Double = 0, m: Int = 0)(implicit vs: MutableCoordinateSpace[T, Double], mult: BinaryOp[H, T, OpMulMatrix, T])
Type Members
-
case class
History(memStep: IndexedSeq[T] = ..., memGradDelta: IndexedSeq[T] = ...) extends Product with Serializable
-
case class
State(iter: Int, initialGNorm: Double, delta: Double, x: T, fval: Double, grad: T, h: H, adjFval: Double, adjGrad: T, history: History) extends Product with Serializable
Value Members
-
final
def
!=(arg0: AnyRef): Boolean
-
final
def
!=(arg0: Any): Boolean
-
final
def
##(): Int
-
final
def
==(arg0: AnyRef): Boolean
-
final
def
==(arg0: Any): Boolean
-
final
def
asInstanceOf[T0]: T0
-
def
chooseDescentDirection(state: State): T
-
def
clone(): AnyRef
-
final
def
eq(arg0: AnyRef): Boolean
-
def
equals(arg0: Any): Boolean
-
def
finalize(): Unit
-
final
def
getClass(): Class[_]
-
def
hashCode(): Int
-
-
final
def
isInstanceOf[T0]: Boolean
-
def
iterations(f: SecondOrderFunction[T, H], initial: T): Iterator[State]
-
lazy val
logger: Logger
-
-
final
def
ne(arg0: AnyRef): Boolean
-
final
def
notify(): Unit
-
final
def
notifyAll(): Unit
-
final
def
synchronized[T0](arg0: ⇒ T0): T0
-
def
toString(): String
-
def
updateHistory(newX: T, newGrad: T, newVal: Double, oldState: State): History
-
final
def
wait(): Unit
-
final
def
wait(arg0: Long, arg1: Int): Unit
-
final
def
wait(arg0: Long): Unit
Inherited from Logging
Inherited from AnyRef
Inherited from Any
Implements a TruncatedNewton Trust region method (like Tron). Also implements "Hessian Free learning". We have a few extra tricks though... :)