Many resources are needed to download a project. Please understand that we have to compensate our server costs. Thank you in advance. Project price only 1 $
You can buy this project and download/modify it how often you want.
package breeze.linalg
import java.util
import breeze.macros.expand
import breeze.linalg.Options._
import breeze.stats.{median, mean}
import scala.reflect.ClassTag
import breeze.math.Semiring
import spire.implicits.{cforRange, cforRange2}
/**
* @author ktakagaki
* @date 04/16/2014.
*/
trait CanPadRight[Input, Dimensions, Output] {
def apply(v: Input, optDim: Dimensions, optMode: OptPadMode): Output
}
object CanPadRight {
//
@expand
implicit def implDV_OptPadDim[@expand.args(Int, Long, Float, Double) T: ClassTag: Semiring]: CanPadRight[DenseVector[T], Dimensions1, DenseVector[T]] =
new CanPadRight[DenseVector[T], Dimensions1, DenseVector[T]] {
def apply(v: DenseVector[T], optDim: Dimensions1, optMode: OptPadMode): DenseVector[T] = {
optMode match {
case Zero => padRight1ImplZero( v, optDim )
case Max => padRight1ImplFixed( v, optDim, max(v) )
case Min => padRight1ImplFixed( v, optDim, min(v) )
case Mean => padRight1ImplFixed( v, optDim, convert( mean(convert(v, Double)), T) ) //option "Mean" with Int will return approximate Int mean for padding....
case Median => padRight1ImplFixed( v, optDim, convert( median(v), T) ) //option "Median" with Int will return approximate Int median for padding....
case Value(n: T) => padRight1ImplFixed( v, optDim, n )
case Wrap => padRight1ImplDV( v, optDim, v )
case Reflect => padRight1ImplDV( v, optDim, reverse(v) )
case _ => throw new IllegalArgumentException("Option " + optMode.toString + " is not supported!")
}
}
def padRight1ImplZero[T]
(v: DenseVector[T], optDim: Dimensions1): DenseVector[T] = {
padRight1ImplFixed(v, optDim, implicitly[Semiring[T]].zero)
}
def padRight1ImplFixed[T](v: DenseVector[T], optDim: Dimensions1, padValue: T): DenseVector[T] = {
require( optDim.n1 > 0, "Cannot pad to zero or negative length!")
v.length match {
case optDim.n1 => v.copy
case num: Int if num < optDim.n1 => DenseVector( v.toArray ++ Array.tabulate(optDim.n1 - num)(p => padValue) )
//ToDo 4: DenseVector.vertcat does not work due to implict problems
//case num: Int if num < optDim.n1 => DenseVector.vertcat( v, DenseVector.tabulate(optDim.n1 - num)(p => padValue) )
case num: Int if optDim.n1 < num => v(0 until optDim.n1).copy //function should return a copy
case _ => throw new IllegalArgumentException("(n) specification incorrect: " + optDim.toString + " !")
}
}
def padRight1ImplDV[T](v: DenseVector[T], optDim: Dimensions1, padDV: DenseVector[T]): DenseVector[T] = {
require( optDim.n1 > 0, "Cannot pad to zero or negative length!")
require( optDim.n1 - v.length <= padDV.length, "Cannot pad beyond specified padding DenseVector!")
v.length match {
case optDim.n1 => v.copy
case num: Int if num < optDim.n1 => DenseVector( v.toArray ++ padDV.toArray.slice(0, optDim.n1 - num) )
//ToDo 4: ditto above
//case num: Int if num < optDim.n1 => DenseVector.vertcat( v, padDV(0 until optDim.n1 - num) )
case num: Int if optDim.n1 < num => v(0 until optDim.n1).copy //function should return a copy
case _ => throw new IllegalArgumentException("(n) specification incorrect: " + optDim.toString + " !")
}
}
}
//
//
@expand
implicit def implDM_OptPadDim_OptPadMode[@expand.args(Int, Long, Float, Double) T: ClassTag: Semiring]: CanPadRight[DenseMatrix[T], Dimensions2, DenseMatrix[T]] =
new CanPadRight[DenseMatrix[T], Dimensions2, DenseMatrix[T]] {
def apply(m: DenseMatrix[T], optDim: Dimensions2, optMode: OptPadMode): DenseMatrix[T] = {
optMode match {
case Zero => padRight2ImplZero( m, optDim )
case Max => padRight2ImplFixed( m, optDim, max(m) )
case Min => padRight2ImplFixed( m, optDim, min(m) )
case Mean => padRight2ImplFixed( m, optDim, convert( mean(convert(m.toDenseVector, Double)), T) )
case Median => padRight2ImplFixed( m, optDim, convert( median(convert(m.toDenseVector, Double)), T) )
case Value(n: T) => padRight2ImplFixed( m, optDim, n )
case Wrap => throw new IllegalArgumentException("Option is not supported for 2D padding.")
case Reflect => throw new IllegalArgumentException("Option is not supported for 2D padding.")
case _ => throw new IllegalArgumentException("Option " + optMode.toString + " is not supported!")
}
}
def padRight2ImplZero[T]
(v: DenseMatrix[T], optDim: Dimensions2): DenseMatrix[T] = {
padRight2ImplFixed(v, optDim, implicitly[Semiring[T]].zero)
}
def padRight2ImplFixed[T](m: DenseMatrix[T], optDim: Dimensions2, padValue: T): DenseMatrix[T] = {
require( optDim.n1 > 0 && optDim.n2 > 0, "Cannot pad to zero or negative length!")
val tempret = DenseMatrix.zeros[T](optDim.n1, optDim.n2)
cforRange2(0 until min(optDim.n2, m.cols), 0 until min(optDim.n1, m.rows)){
(c, r) => tempret(r, c) = m(r, c)
}
tempret
}
}
//
}
trait CanPadLeft[Input, Dimensions, Output] {
def apply(v: Input, optDim: Dimensions, optMode: OptPadMode): Output
}
object CanPadLeft {
//
@expand
implicit def implDV_OptPadDim[@expand.args(Int, Long, Float, Double) T: ClassTag: Semiring]: CanPadLeft[DenseVector[T], Dimensions1, DenseVector[T]] =
new CanPadLeft[DenseVector[T], Dimensions1, DenseVector[T]] {
def apply(v: DenseVector[T], optDim: Dimensions1, optMode: OptPadMode): DenseVector[T] = {
optMode match {
case Zero => padLeft1ImplZero( v, optDim )
case Max => padLeft1ImplFixed( v, optDim, max(v) )
case Min => padLeft1ImplFixed( v, optDim, min(v) )
case Mean => padLeft1ImplFixed( v, optDim, convert( mean(convert(v, Double)), T) ) //option "Mean" with Int will return approximate Int mean for padding....
case Median => padLeft1ImplFixed( v, optDim, convert( median(v), T) ) //option "Median" with Int will return approximate Int median for padding....
case Value(n: T) => padLeft1ImplFixed( v, optDim, n )
case Wrap => padLeft1ImplDV( v, optDim, v )
case Reflect => padLeft1ImplDV( v, optDim, reverse(v) )
case _ => throw new IllegalArgumentException("Option " + optMode.toString + " is not supported!")
}
}
def padLeft1ImplZero
(v: DenseVector[T], optDim: Dimensions1): DenseVector[T] = {
padLeft1ImplFixed(v, optDim, implicitly[Semiring[T]].zero)
}
def padLeft1ImplFixed(v: DenseVector[T], optDim: Dimensions1, padValue: T): DenseVector[T] = {
require( optDim.n1 > 0, "Cannot pad to zero or negative length!")
v.length match {
case optDim.n1 => v.copy
case num: Int if num < optDim.n1 =>
val res = new Array[T](optDim.n1)
util.Arrays.fill(res, padValue)
val r = DenseVector(res)
r((optDim.n1 - num) until optDim.n1) := v
r
case num: Int if optDim.n1 < num => v(v.length - optDim.n1 until v.length).copy //function should return a copy
case _ => throw new IllegalArgumentException("(n) specification incorrect: " + optDim.toString + " !")
}
}
def padLeft1ImplDV[T](v: DenseVector[T], optDim: Dimensions1, padDV: DenseVector[T]): DenseVector[T] = {
require( optDim.n1 > 0, "Cannot pad to zero or negative length!")
require( optDim.n1 - v.length <= padDV.length, "Cannot pad beyond specified padding DenseVector!")
v.length match {
case optDim.n1 => v.copy
case num: Int if num < optDim.n1 => DenseVector( reverse(reverse(padDV).apply(0 until optDim.n1 - num)).toArray ++ v.toArray )
case num: Int if optDim.n1 < num => v(0 until optDim.n1).copy //function should return a copy
case _ => throw new IllegalArgumentException("(n) specification incorrect: " + optDim.toString + " !")
}
}
}
//
//
@expand
implicit def implDM_OptPadDim_OptPadMode[@expand.args(Int, Long, Float, Double) T: ClassTag: Semiring]: CanPadLeft[DenseMatrix[T], Dimensions2, DenseMatrix[T]] =
new CanPadLeft[DenseMatrix[T], Dimensions2, DenseMatrix[T]] {
def apply(m: DenseMatrix[T], optDim: Dimensions2, optMode: OptPadMode): DenseMatrix[T] = {
optMode match {
case Zero => padLeft2ImplZero( m, optDim )
case Max => padLeft2ImplFixed( m, optDim, max(m) )
case Min => padLeft2ImplFixed( m, optDim, min(m) )
case Mean => padLeft2ImplFixed( m, optDim, convert( mean(convert(m.toDenseVector, Double)), T) )
case Median => padLeft2ImplFixed( m, optDim, convert( median(convert(m.toDenseVector, Double)), T) )
case Value(n: T) => padLeft2ImplFixed( m, optDim, n )
case Wrap => throw new IllegalArgumentException("Option is not supported for 2D padding.")
case Reflect => throw new IllegalArgumentException("Option is not supported for 2D padding.")
case _ => throw new IllegalArgumentException("Option " + optMode.toString + " is not supported!")
}
}
def padLeft2ImplZero[T]
(v: DenseMatrix[T], optDim: Dimensions2): DenseMatrix[T] = {
padLeft2ImplFixed(v, optDim, implicitly[Semiring[T]].zero)
}
def padLeft2ImplFixed[T](m: DenseMatrix[T], optDim: Dimensions2, padValue: T): DenseMatrix[T] = {
require( optDim.n1 > 0 && optDim.n2 > 0, "Cannot pad to zero or negative length!")
val tempret = DenseMatrix.zeros[T](optDim.n1, optDim.n2)
cforRange2(1 to min(optDim.n2, m.cols), 1 to min(optDim.n1, m.rows)) {
(c, r) => tempret(optDim.n1 - r , optDim.n2 - c) = m(m.rows - r, m.cols - c)
}
tempret
}
}
//
}
