|
|
| (One intermediate revision by one other user not shown) |
| Line 1: |
Line 1: |
| In [[linear algebra]], a '''Householder transformation''' (also known as '''Householder reflection''' or '''elementary reflector''') is a [[linear transformation]] that describes a [[reflection (mathematics)|reflection]] about a [[plane (mathematics)|plane]] or [[hyperplane]] containing the origin. Householder transformations are widely used in [[numerical linear algebra]], to perform [[QR decomposition]]s and in the first step of the [[QR algorithm]]. The Householder transformation was introduced in 1958 by [[Alston Scott Householder]].<ref>{{cite journal
| | My name: Laurel Northern<br>Age: 36 years old<br>Country: Brazil<br>Home town: Joinville <br>Postal code: 89237-415<br>Address: Rua Jacob Moser 851<br><br>Take a look at my web blog :: [http://www.chrisp-voc.de/gaestebuch.html?180%FFm88%FFnofollow%FFSims+4+Crack%FFnofollow%FFsuperfit+cla+fat+burner+review%FF%FFThe+Sims+4+Crack%FF%FFchrisp-voc.de%FFexternal+nofollow fifa 15 coin Generator] |
| |first=A. S. |last=Householder |authorlink=Alston Scott Householder
| |
| |title=Unitary Triangularization of a Nonsymmetric Matrix
| |
| |journal=[[Journal of the ACM]]
| |
| |volume=5 |issue=4 |year=1958 |pages=339–342
| |
| |doi=10.1145/320941.320947 |mr=0111128
| |
| }}</ref>
| |
| | |
| Its analogue over general [[inner product spaces]] is the [[Householder operator]].
| |
| | |
| ==Definition and properties==
| |
| | |
| The reflection hyperplane can be defined by a [[unit vector]] ''v'' (a vector with length 1) which is [[orthogonal]] to the hyperplane. The reflection of a [[Point (geometry)|point]] ''x'' about this hyperplane is:
| |
| | |
| :<math>x - 2\langle v,x\rangle v = x - 2 v (v^\text{H} x), </math> | |
| | |
| where ''v'' is given as a column unit vector with [[Hermitian transpose]] ''v''<sup>H</sup>. This is a [[linear transformation]] given by the '''Householder matrix''':
| |
| | |
| : <math>P = I - 2 v v^\text{H}\,</math>, where ''I'' is the [[identity matrix]].
| |
| | |
| The Householder matrix has the following properties:
| |
| * it is [[Hermitian matrix|Hermitian]]: <math>P = P^\text{H}, </math>
| |
| * it is [[Unitary matrix|unitary]]: <math>P^{-1}=P^\text{H}, </math>
| |
| * hence it is [[involution (mathematics)|involutary]]: <math>P^2=I </math>.
| |
| * A Householder matrix has eigenvalues <math>\pm 1</math>. To see this, notice that if <math>u</math> is orthogonal to the vector <math>v</math> which was used to create the reflector, then <math>Pu = u</math>, i.e., 1 is an eigenvalue of multiplicity <math>n-1</math>, since there are <math>n-1</math> independent vectors orthogonal to <math>v</math>. Also, notice <math>Pv = -v</math>, and so -1 is an eigenvalue with multiplicity 1.
| |
| * The determinant of a Householder reflector is -1, since the determinant of a matrix is the product of its eigenvalues.
| |
| | |
| == Applications ==
| |
| In geometric optics, [[specular reflection]] can be expressed in terms of the Householder matrix.
| |
| | |
| Householder reflections can be used to calculate [[QR decomposition]]s by reflecting first one column of a matrix onto a multiple of a standard basis vector, calculating the transformation matrix, multiplying it with the original matrix and then recursing down the (''i'', ''i'') [[minor (linear algebra)|minor]]s of that product.
| |
| | |
| They are also widely used for [[tridiagonal]]ization of symmetric matrices and for transforming non-symmetric matrices to a [[Hessenberg matrix|Hessenberg]] form.
| |
| | |
| === Tridiagonalization ===
| |
| {{main|Tridiagonal matrix}}
| |
| This procedure is taken from the book: Numerical Analysis, Burden and Faires, 8th Edition.
| |
| In the first step, to form the Householder matrix in each step we need to determine <math> \displaystyle \alpha</math> and r, which are:
| |
| | |
| :<math> \displaystyle \alpha = -\operatorname{sgn}(a_{21})\sqrt{\sum_{j=2}^{n}a_{j1}^2} </math>; | |
| | |
| :<math> r = \sqrt{\frac{1}{2}(\alpha^{2}-a_{21}\alpha)} </math>;
| |
| | |
| From <math> \displaystyle \alpha</math> and r, construct vector v:
| |
| | |
| :<math> v^{(1)} = \begin{bmatrix} v_1 \\ v_2 \\ \vdots \\ v_n \end{bmatrix},</math>
| |
| | |
| where <math>v_1=0;</math>, <math> v_2 = \frac{a_{21}-\alpha}{2r}</math>, and
| |
| :<math> v_k = \frac{a_{k1}}{2r}</math> for each k=3,4 ..n
| |
| | |
| Then compute:
| |
| | |
| :<math> \displaystyle P^{1} = I - 2v^{(1)}(v^{(1)})^\text{T}</math>
| |
| | |
| :<math>\displaystyle A^{(2)} = P^{1}AP^{1}</math>
| |
| | |
| Having found <math> \displaystyle P^{1}</math> and computed <math>\displaystyle A^{(2)}</math> the process is repeated for k =2, 3, ..., ''n-1'' as follows:
| |
| | |
| :<math> \displaystyle \alpha = -\operatorname{sgn}(a^k_{k+1,k})\sqrt{\sum_{j=k+1}^{n}(a^k_{jk})^2} </math>;
| |
| | |
| :<math> r = \sqrt{\frac{1}{2}(\alpha^{2}-a^k_{k+1,k}\alpha)} </math>;
| |
| | |
| :<math>v^k_1 = v^{k}_2 = \cdots = v^k_k=0;</math>
| |
| | |
| :<math> v^{k}_{k+1} = \frac{a^{k}_{k+1,k}-\alpha}{2r}</math>
| |
| | |
| :<math> v^{k}_j = \frac{a^{k}_{jk}}{2r}</math> for j = k + 2; k + 3, ..., n
| |
| | |
| :<math> \displaystyle P^{k} = I - 2v^{(k)}(v^{(k)})^\text{T}</math>
| |
| | |
| :<math>\displaystyle A^{(k+1)} = P^{k}A^{(k)}P^{k}</math>
| |
| | |
| Continuing in this manner, the tridiagonal and symmetric matrix is formed.
| |
| | |
| ===Examples===
| |
| | |
| This example is taken from the book "Numerical Analysis" by Richard L. Burden (Author), J. Douglas Faires. In this example, the given matrix is transformed to the similar tridiagonal matrix A<sub>2</sub> by using Householder Method.
| |
| | |
| <math>\mathbf{A} = \begin{bmatrix}
| |
| | |
| 4&1&-2&2 \\ | |
| 1 & 2 &0&1 \\
| |
| -2 & 0 &3& -2 \\
| |
| 2 & 1 & -2&-1 \end{bmatrix},</math>
| |
| | |
| Following those steps in Householder Method. We have:
| |
| | |
| The first Householder matrix:
| |
| | |
| Q<sub>1</sub>
| |
| <math>\mathbf{} = \begin{bmatrix}
| |
| | |
| 1&0&0&0 \\
| |
| 0 &-1/3&2/3&-2/3 \\
| |
| 0 & 2/3 &2/3& 1/3 \\
| |
| 0 & -2/3 &1/3& 2/3 \end{bmatrix},</math>
| |
| | |
| A<sub>1</sub> = Q<sub>1</sub>AQ<sub>1</sub> = <math>\mathbf{}\begin{bmatrix}
| |
| | |
| 4&-3&0&0 \\
| |
| -3 & 10/3 &1&4/3 \\
| |
| 0 & 1 &5/3& -4/3 \\
| |
| 0 & 4/3 & -4/3&-1 \end{bmatrix},</math>
| |
| | |
| Used A<sub>1</sub> to form Q<sub>2</sub> =<math>\mathbf{}\begin{bmatrix}
| |
| | |
| 1&0&0&0 \\
| |
| 0&1 &0&0 \\
| |
| 0 & 0 &-3/5&-4/5 \\
| |
| 0 & 0 & -4/5&3/5 \end{bmatrix},</math>
| |
| | |
| A<sub>2</sub> = Q<sub>2</sub>A<sub>1</sub>Q<sub>2</sub>=<math>\mathbf{}\begin{bmatrix}
| |
| | |
| 4&-3&0&0 \\
| |
| -3 &10/3 &-5/3&0 \\
| |
| 0 & -5/3 &-33/25& 68/75 \\
| |
| 0 &0 & 68/75&149/75 \end{bmatrix},</math>
| |
| | |
| As we can see, the final result is a tridiagonal symmetric matrix which is similar to the original one. The process finished after 2 steps.
| |
| | |
| == Computational and Theoretical Relationship to other Unitary Transformations ==
| |
| | |
| {{see also|Rotation (mathematics)}}
| |
| The Householder Transformation is a reflection about a certain hyperplane, namely, the one with unit normal vector v, as stated earlier. An N by N [[unitary transformation]] U satisfies ''UU''<sup>H</sup>=''I''. Taking determinant (N-th power of the geometric mean) and trace (proportional to arithmetic mean) of a unitary matrix reveals that its eigenvalues λ<sub>i</sub> are unit modulus. This can be seen directly and swiftly:
| |
| :<math> \frac{\mbox{Trace}(UU^\text{H})}{N}=\frac{\sum_{j=2}^N|\lambda_j|^2}{N}=1, \mbox{det}(UU^\text{H})=\prod_{j=1}^N |\lambda_j|^2=1. </math>
| |
| Since arithmetic and geometric means are equal iff the variables are constant, see, [[inequality of arithmetic and geometric means]], we establish the claim of unit modulus.
| |
| | |
| For the case of real valued unitary matrixes we obtain [[orthogonal matrices]], <math> U U^\text{T}=I. </math> In this case all eigenvalues are real, and so the unit modulus eigenvalue constraint is replaced by the binary constraint that all eigenvalues lie in the set {+1,-1}.
| |
| It follows rather readily (see [[orthogonal matrix]]) that any orthogonal matrix can be [[QR_decomposition#Using_Givens_rotations|decomposed]] into a product of 2 by 2 rotations, called [[Givens rotation|Givens Rotations]], and Householder reflections. This is appealing intuitively since multiplication of a vector by an orthogonal matrix preserves the length of that vector, and rotations and reflections exhaust the set of (real valued) geometric operations that render invariant a vector's length.
| |
| | |
| The Householder transformation was shown to have a one to one relationship with the canonical coset decomposition of unitary matrices defined in group theory, which can be used to parametrize unitary operators in a very efficient manner.<ref>{{cite journal
| |
| |first= |last= Renan Cabrera, Traci Strohecker, and Herschel Rabitz
| |
| |title= The canonical coset decomposition of unitary matrices through Householder transformations
| |
| |journal=[[Journal of Mathematical Physics]]
| |
| |volume=51 |issue=8 |year=2010
| |
| |doi=10.1063/1.3466798
| |
| }}</ref>
| |
| | |
| Finally we note that a single Householder Transform, unlike a solitary Givens Transform, can act on all columns of a matrix, and as such exhibits the lowest computational cost for QR decomposition and Tridiagonalization. The penalty for this "computational optimality" is, of course, that Householder operations cannot be as deeply or efficiently parallelized. As such Householder is preferred for dense matrices on sequential machines, whilst Givens is preferred on sparse matrices, and/or parallel machines.
| |
| | |
| ==References==
| |
| <references />
| |
| * {{cite journal
| |
| |first=C.D. |last=LaBudde
| |
| |title=The reduction of an arbitrary real square matrix to tridiagonal form using similarity transformations
| |
| |journal=[[Mathematics of Computation]]
| |
| |volume=17 |issue=84 |year=1963 |pages=433–437
| |
| |mr=0156455 |doi=10.2307/2004005
| |
| |jstor=2004005
| |
| |publisher=American Mathematical Society
| |
| }}
| |
| * {{cite journal
| |
| |first=D.D. |last=Morrison
| |
| |title=Remarks on the Unitary Triangularization of a Nonsymmetric Matrix
| |
| |journal=[[Journal of the ACM]]
| |
| |volume=7 |issue=2 |year=1960 |pages=185–186
| |
| |doi=10.1145/321021.321030 |mr=0114291
| |
| }}
| |
| | |
| * {{cite journal
| |
| |first=Barry |last=Cipra
| |
| |title=The Best of the 20th Century: Editors Name Top 10 Algorithms
| |
| |volume=33 | issue=4 | year= 2000 | page= 1| url=http://www.siam.org/pdf/news/637.pdf}} (Herein Householder Transformation is cited as a top 10 algorithm of this century)
| |
| | |
| *{{Cite book | last1=Press | first1=WH | last2=Teukolsky | first2=SA | last3=Vetterling | first3=WT | last4=Flannery | first4=BP | year=2007 | title=Numerical Recipes: The Art of Scientific Computing | edition=3rd | publisher=Cambridge University Press | publication-place=New York | isbn=978-0-521-88068-8 | chapter=Section 11.3.2. Householder Method | chapter-url=http://apps.nrbook.com/empanel/index.html#pg=578 | postscript=<!-- Bot inserted parameter. Either remove it; or change its value to "." for the cite to end in a ".", as necessary. -->{{inconsistent citations}}}}
| |
| | |
| == External links ==
| |
| *[http://math.fullerton.edu/mathews/n2003/HouseholderMod.html Householder Transformations]
| |
| | |
| {{Numerical linear algebra}}
| |
| | |
| [[Category:Transformation (function)]]
| |
| [[Category:Matrices]]
| |
| [[Category:Numerical linear algebra]]
| |