问题解决了。请在本StackOverflow问题中查看我自己的答案,了解具体方法。
但是,这是新的(并且正常工作)代码:
显示器与下面相同。
/**
* Returns the identity matrix of the specified dimension
* @param size the number of columns (i.e. the number of rows) of the desired identity matrix
* @return the identity matrix of the specified dimension
*/
def getIdentityMatrix(size : Int): scala.collection.mutable.Seq[scala.collection.mutable.Seq[Double]] = {
scala.collection.mutable.Seq.tabulate(size)(r => scala.collection.mutable.Seq.tabulate(size)(c => if(r == c) 1.0 else 0.0))
}
/**
* This algorithm processes column by column.
* STEP 1. It finds the greatest coefficient for the current column (called 'a') and, if it equals 0, returns NULL (since the matrix
* can't be inverted) ; otherwise (STEP 2.), it swaps the pivot's line with this new line and the pivot becomes the adequate coefficient
* of this new line
* STEP 3. It divides the pivot's line by the pivot
* STEP 4. It sets each of the current column's coefficient to 0 by subtracting the corresponding lines by the pivot's line
* @return
*/
def getGaussJordanInvertedMatrix: (Matrix, Matrix) = {
// We get first the matrix to be inverted, second the identity one
val mutable_being_inversed_matrix : collection.mutable.Seq[collection.mutable.Seq[Double]] = scala.collection.mutable.Seq(content.map(ms => scala.collection.mutable.Seq(ms:_*)):_*)
val identity_matrix : collection.mutable.Seq[collection.mutable.Seq[Double]] = getIdentityMatrix(content.length) // We get the identity matrix. It will be modified
// as the original matrix will.
var id_last_pivot : Int = 0 // ID of the last pivot, i.e. ID of the current column
content.indices.foreach(general_id_column => {
println("Current column : " + general_id_column)
// STEP 1.
val id_line_with_max_coefficient_in_this_column = (id_last_pivot until content.length).maxBy(id_line_in_this_column => Math.abs(mutable_being_inversed_matrix(id_line_in_this_column)(general_id_column)))
if(mutable_being_inversed_matrix(id_line_with_max_coefficient_in_this_column)(general_id_column) == 0) {
println("The Gauss-Jordan elimination's algorithm returns an error : indeed, the matrix can't be inverted")
} else {
// STEP 2.
val tmp_line : scala.collection.mutable.Seq[Double] = mutable_being_inversed_matrix(id_last_pivot)
mutable_being_inversed_matrix(id_last_pivot) = mutable_being_inversed_matrix(id_line_with_max_coefficient_in_this_column)
mutable_being_inversed_matrix(id_line_with_max_coefficient_in_this_column) = tmp_line
val identity_tmp_line : scala.collection.mutable.Seq[Double] = identity_matrix(id_last_pivot)
identity_matrix(id_last_pivot) = identity_matrix(id_line_with_max_coefficient_in_this_column)
identity_matrix(id_line_with_max_coefficient_in_this_column) = identity_tmp_line
println("\nSWAP DONE")
println(Console.BLUE + "Original matrix :\n" + Console.RESET + mutable_being_inversed_matrix.mkString("\n"))
println(Console.RED + "Identity matrix :\n" + Console.RESET + identity_matrix.mkString("\n"))
// STEP 3.
val tmp = mutable_being_inversed_matrix(id_last_pivot)(general_id_column)
mutable_being_inversed_matrix(id_last_pivot) = mutable_being_inversed_matrix(id_last_pivot).map(coefficient => coefficient / tmp)
identity_matrix(id_last_pivot) = identity_matrix(id_last_pivot).map(coefficient => coefficient / tmp)
println("\nDIVISION DONE")
println(Console.BLUE + "Original matrix :\n" + Console.RESET + mutable_being_inversed_matrix.mkString("\n"))
println(Console.RED + "Identity matrix :\n" + Console.RESET + identity_matrix.mkString("\n"))
// STEP 4.
content.indices.foreach(id_line => {
val tmp = mutable_being_inversed_matrix(id_line)(general_id_column)
if(id_line != id_last_pivot) {
content.indices.foreach(id_column => {
mutable_being_inversed_matrix(id_line)(id_column) -= mutable_being_inversed_matrix(id_last_pivot)(id_column) * tmp
identity_matrix(id_line)(id_column) -= identity_matrix(id_last_pivot)(id_column) * tmp
})
}
})
println("\nSUBTRACTION & MULTIPLICATION DONE")
println(Console.BLUE + "Original matrix :\n" + Console.RESET + mutable_being_inversed_matrix.mkString("\n"))
println(Console.RED + "Identity matrix :\n" + Console.RESET + identity_matrix.mkString("\n"))
println()
id_last_pivot += 1
}
})
(new Matrix(identity_matrix), new Matrix(mutable_being_inversed_matrix))
}
我试图实现高斯 - 乔丹消除的Scala版本来反转矩阵(NB:可变集合和命令范式用于简化实现 - 我尝试编写算法,但它&#&# 39;几乎不可能,因此算法包含嵌套步骤。
单位矩阵没有很好地转化为反演的结果。换句话说:将单位矩阵转换为倒置矩阵(这是Gauss-Jordan消除的结果)是不正确的。
考虑这个矩阵(A):
(2.0,-1.0,0.0)
( - 1.0,2.0,-1.0)
(0.0,-1.0,2.0)
这一个(B):
(1.0,0.0,0.0)
(0.0,1.0,0.0)
(0.0,0.0,1.0)
如果我们应用Gauss-Jordan消除,A变为:
(1.0,0.0,0.0)
(0.0,1.0,0.0)
(0.0,0.0,1.0)
如果我们应用Gauss-Jordan消除,B变为:
(0.75 0.5 0.25)
(0.5 1 0.5)
(0.25 0.5 0.75)
如果我们应用我的实现,A就没有问题,因为我得到了以下矩阵:
(1.0,0.0,0.0)
(0.0,1.0,0.0)
(0.0,0.0,1.0)
但是,如果我们应用我的实现,B没有很好地转换,因为我获得了以下矩阵:
(1.0,0.5,0.0)
(1.0,0.5,0.66666666666666666)
(0.0,1.0,0.33333333333333337)
逐行进行3个步骤。这些步骤是:
^ 2:当前列中的最大系数是从第(z + 1)行找到的,其中z是我们使用的最后一个数据透视表的ID(即:最后一个工作列的ID)
我们将包含我们在步骤1得到的枢轴的整行划分为枢轴,将枢轴设置为1(在后面的句子中,表达式"枢轴"系统地指的是此枢轴我们得到了STEP 1)。顺便说一下,注意不太重要的事实,即同一条线的其他系数也是分开的(参见"我们将整行划分为#34;)。
我们将当前列的每一行整数减去绕轴的行,将所有当前列的系数设置为0.顺便说一下,注意不太重要的事实这些相同行的其他系数也被减去(参见"我们减去每一行")。
val m : Matrix = new Matrix(Seq(Seq(2, -1, 0), Seq(-1, 2, -1), Seq(0, -1, 2)))
val m : Matrix = new Matrix(Seq(Seq(2, -1, 0), Seq(-1, 2, -1), Seq(0, -1, 2)))
println("ORIGINAL MATRIX =\n" + m)
println
val result : (Matrix, Matrix) = m.getGaussJordanInvertedMatrix
println()
println("RESULT =\n" + Console.BLUE + "Original matrix :\n" + Console.RESET + result._2 + Console.RED + "\nIdentity matrix :\n" + Console.RESET + result._1)
/**
* Returns the identity matrix of the specified dimension
* @param size the number of columns (i.e. the number of rows) of the desired identity matrix
* @return the identity matrix of the specified dimension
*/
def getIdentityMatrix(size : Int): scala.collection.mutable.Seq[scala.collection.mutable.Seq[Double]] = {
scala.collection.mutable.Seq.tabulate(size)(r => scala.collection.mutable.Seq.tabulate(size)(c => if(r == c) 1.0 else 0.0))
}
/**
* This algorithm processes column by column.
* STEP 1. It finds the greatest coefficient for the current column (called 'a') and, if it equals 0, returns NULL (since the matrix
* can't be inverted) ; otherwise (STEP 2.), it swaps the pivot's line with this new line and the pivot becomes the adequate coefficient
* of this new line
* STEP 3. It divides the pivot's line by the pivot
* STEP 4. It sets each of the current column's coefficient to 0 by subtracting the corresponding lines by the pivot's line
* @return
*/
def getGaussJordanInvertedMatrix: (Matrix, Matrix) = {
// We get first the matrix to be inverted, second the identity one
val mutable_being_inversed_matrix : collection.mutable.Seq[collection.mutable.Seq[Double]] = scala.collection.mutable.Seq(content.map(ms => scala.collection.mutable.Seq(ms:_*)):_*)
val identity_matrix : collection.mutable.Seq[collection.mutable.Seq[Double]] = getIdentityMatrix(content.length) // We get the identity matrix. It will be modified
// as the original matrix will.
var id_last_pivot : Int = 0 // ID of the last pivot, i.e. ID of the current column
content.indices.foreach(general_id_column => {
println("Current column : " + general_id_column)
// STEP 1.
val id_line_with_max_coefficient_in_this_column = (id_last_pivot until content.length).maxBy(id_line_in_this_column => Math.abs(mutable_being_inversed_matrix(id_line_in_this_column)(general_id_column)))
if(mutable_being_inversed_matrix(id_line_with_max_coefficient_in_this_column)(general_id_column) == 0) {
println("The Gauss-Jordan elimination's algorithm returns an error : indeed, the matrix can't be inverted")
} else {
// STEP 2.
val tmp_line : scala.collection.mutable.Seq[Double] = mutable_being_inversed_matrix(id_last_pivot)
mutable_being_inversed_matrix(id_last_pivot) = mutable_being_inversed_matrix(id_line_with_max_coefficient_in_this_column)
mutable_being_inversed_matrix(id_line_with_max_coefficient_in_this_column) = tmp_line
val identity_tmp_line : scala.collection.mutable.Seq[Double] = identity_matrix(id_last_pivot)
identity_matrix(id_last_pivot) = identity_matrix(id_line_with_max_coefficient_in_this_column)
identity_matrix(id_line_with_max_coefficient_in_this_column) = identity_tmp_line
println("\nSWAP DONE")
println(Console.BLUE + "Original matrix :\n" + Console.RESET + mutable_being_inversed_matrix.mkString("\n"))
println(Console.RED + "Identity matrix :\n" + Console.RESET + identity_matrix.mkString("\n"))
// STEP 3.
mutable_being_inversed_matrix(id_last_pivot) = mutable_being_inversed_matrix(id_last_pivot).map(coefficient => coefficient / mutable_being_inversed_matrix(id_last_pivot)(general_id_column))
identity_matrix(id_last_pivot) = identity_matrix(id_last_pivot).map(coefficient => coefficient / mutable_being_inversed_matrix(id_last_pivot)(general_id_column))
println("\nDIVISION DONE")
println(Console.BLUE + "Original matrix :\n" + Console.RESET + mutable_being_inversed_matrix.mkString("\n"))
println(Console.RED + "Identity matrix :\n" + Console.RESET + identity_matrix.mkString("\n"))
// STEP 4.
content.indices.foreach(id_line => {
val tmp = mutable_being_inversed_matrix(id_line)(general_id_column)
if(id_line != id_last_pivot) {
content.indices.foreach(id_column => {
mutable_being_inversed_matrix(id_line)(id_column) -= mutable_being_inversed_matrix(id_last_pivot)(id_column) * tmp
identity_matrix(id_line)(id_column) -= mutable_being_inversed_matrix(id_last_pivot)(id_column) * tmp
})
}
})
println("\nSUBTRACTION & MULTIPLICATION DONE")
println(Console.BLUE + "Original matrix :\n" + Console.RESET + mutable_being_inversed_matrix.mkString("\n"))
println(Console.RED + "Identity matrix :\n" + Console.RESET + identity_matrix.mkString("\n"))
println()
id_last_pivot += 1
}
})
(new Matrix(identity_matrix), new Matrix(mutable_being_inversed_matrix))
}
您可以在此处使用此输入找到我的实施的执行:https://jsfiddle.net/wwhdu32x/
您可以在此处找到疑难解答:https://jsfiddle.net/wwhdu32x/1/(以&#34开头的评论; ERROR"写入 - 注意:此故障排除仅涉及第一次迭代,即第一列)。
为什么我的身份矩阵没有得到很好的转变?我怎么能处理它?</ p>
答案 0 :(得分:0)
问题解决了。问题已经更新,其中包括新代码(旧代码仍然可用,以便进行比较)。有两个错误(下面的“STEP XYZ”引用了相应的源代码的STEP,而不是这个StackOverflow问题中提到的步骤,它们有点不同):
关于单位矩阵的减法不使用单位矩阵的系数(步骤4)。错误修复:identity_matrix(id_line)(id_column) -= identity_matrix(id_last_pivot)(id_column) * tmp
其次,在STEP 3中,我忘了将枢轴存储在临时变量中,以便将两个矩阵(原始矩阵和标识矩阵)与它分开。在不存储它的情况下,在原始矩阵上划分后,枢轴的值发生了变化。错误修复:
val tmp = mutable_being_inversed_matrix(id_last_pivot)(general_id_column)
mutable_being_inversed_matrix(id_last_pivot) = mutable_being_inversed_matrix(id_last_pivot).map(coefficient => coefficient / tmp)
identity_matrix(id_last_pivot) = identity_matrix(id_last_pivot).map(coefficient => coefficient / tmp)