{"id":652,"date":"2018-11-03T14:40:46","date_gmt":"2018-11-03T14:40:46","guid":{"rendered":"https:\/\/konrad.burnik.org\/wordpress\/?p=652"},"modified":"2024-08-04T20:28:28","modified_gmt":"2024-08-04T20:28:28","slug":"centrosymmetric-qr-decomposition-and-matrices-of-split-complex-numbers","status":"publish","type":"post","link":"https:\/\/konrad.burnik.org\/wordpress\/centrosymmetric-qr-decomposition-and-matrices-of-split-complex-numbers\/","title":{"rendered":"Centrosymmetric QR decomposition and its connection to split-complex matrices"},"content":{"rendered":"

<\/p>\n

Introduction<\/h2>\n

Back in 2015 I published an article<\/a> [1] about the QR decomposition<\/a> of a centrosymmetric real matrix. In 2016 I started thinking about the meaning of this decomposition and centrosymmetric matrices in general. I discovered that centrosymmetric matrices have a natural interpretation as split-complex matrices and even more, that the centrosymmetric QR decomposition of a matrix actually corresponds to a QR decompositon of a split-complex matrix for which the original matrix is it\u00a0representation<\/em>. In a certain sense, the centrosymmetric QR decomposition introduced in [1] of a real square matrix of order \"2n\" is equivalent to a QR decomposition of a corresponding split-complex square matrix of order \"n\".\u00a0 All these notions will be made precise in the following sections. This blog post is based on my own\u00a0preprint<\/a>.<\/p>\n

Matrix representations<\/h2>\n

If \"V\" is a finite-dimensional vector space then we denote by \"\mathcal{L}(V)\" the set of all linear transformations \"V \rightarrow V\". Recall that if \"V\" is an \"n\"-dimensional vector space and \"\mathcal{B} = (v_1,\dots,v_n)\" an ordered basis for \"V\", then every linear transformation \"T : V \rightarrow V\" has an \"n \times n\" matrix representation with respect to \"\mathcal{B}\" denoted \"[T]_\mathcal{B}\". Further, for any two linear transformations \"S, T: V \rightarrow V\" we have \"[S \circ T]_\mathcal{B} = [S]_\mathcal{B} [T]_\mathcal{B}\". The standard ordered basis for \"\mathbb{R}^n\" i.e. the basis \"(e_1,\dots,e_n)\" is defined as \"e_{i,j} = 1\" if \"i=j\" and \"e_{i,j} =0\" otherwise.<\/p>\n

An algebra<\/em> \"\mathfrak{A}\" is an ordered pair \"(A, \cdot)\" such that \"A\" is a vector space over a field \"K\" and \"\cdot: A \times A \rightarrow A\" is a bilinear mapping called multiplication<\/em>.<\/p>\n

Let \"\mathfrak{A} = (A, \cdot)\" be an algebra. A representation<\/em> of \"\mathfrak{A}\" over a vector space \"V\" is a map \"\phi : A \rightarrow \mathcal{L}(V)\" such that<\/p>\n

\"\phi(a_1 \cdot a_2) = \phi(a_1) \circ \phi(a_2)\" for all \"a_1, a_2 \in A\".<\/p>\n

Let \"M_n(\mathbb{R})\" denote the set of \"n \times n\" real matrices. If \"V\" is an \"n\"-dimensional vector space and \"\mathcal{B} = (v_1,\dots,v_n)\" an ordered basis for \"V\" then every linear transformation \"T : V\rightarrow V\" has a matrix representation \"[T]_\mathcal{B} \in M_n(\mathbb{R})\". For each \"a \in A\" we have \"\phi(a) \in \mathcal{L}(V)\". Since \"V\" is \"n\"-dimensional, we have and ordered basis \"\mathcal{B}\" and \"[\phi(a)]_\mathcal{B} \in M_n(\mathbb{R})\". A matrix representation<\/em> of \"\mathfrak{A}\" with respect to \"\mathcal{B}\"} is a map \"\phi_\mathcal{B} : A \rightarrow M_n(\mathbb{R})\" such that \"\phi_\mathcal{B}(a) = [\phi(a)]_\mathcal{B}\" for all \"a \in A\". Further, we have \"\phi_\mathcal{B}(a_1 \cdot a_2) = \phi(a_1)_\mathcal{B}\cdot \phi(a_2)_\mathcal{B}\" for all \"a_1,a_2 \in A\". These are all well known notions from representation theory, for further information, one can consult one of the standard textbooks, for example see [3].<\/p>\n

Algebra of split-complex numbers<\/h2>\n

Let \"\Cs = (\mathbb{R}^2, \cdot)\", where \"\cdot : \mathbb{R}^2 \times \mathbb{R}^2 \rightarrow \mathbb{R}^2\" is given by
<\/a>

(1) <\/span>   <\/span>\"\begin{align*}(a, b) \cdot (c, d) &= (ac + bd, ad + bc)\end{align*}\"<\/p> for all \"(a, b), (c, d) \in \mathbb{R}^2\". It is straightforward to verify that \"\mathfrak{D}\" is
an algebra. This is the well known algebra of\u00a0 split-complex numbers<\/em>. The split-complex numbers are also sometimes known as hyperbolic numbers<\/em>.
Similarly as for the complex numbers, each real number \"x \in \mathbb{R}\" can be identified with the pair \"(x, 0)\".
With this correspondence, the pair \"j \equiv (0, 1)\" has the property \"j^2 = +1\" and \"j \not = \pm 1\". Due to this property, \"j\" is called the hyperbolic unit<\/em>.
Since \"(x, y) = (x, 0) + (0, 1) \cdot (0, y)\", in the following we shall denote a pair \"(x, y)\" simply with \"x + j y\".\u00a0The conjugate<\/em> of \"z = a + j b\" is defined as \"z^* = a - j b\".<\/p>\n

For a hyperbolic number \"z = a + j b\" we define the real part<\/em> as \"\Real(z) = a\" and hyperbolic part<\/em> as \"\Hyp(z) = b\".
For the module we set \"|z| = zz^* = a^2 - b^2\" and we have \"|w \cdot z| = |w|\cdot|z|\" for all \"w, z \in \Cs\".
For an extensive overview of the theory of hyperbolic numbers as well of their usefulness in Physics one can check the literature, for example [2].\u00a0For the rest of this blog post, we shall refer to these numbers as split-complex numbers<\/em>.<\/p>\n

Centrosymmetric representation of split-complex matrices<\/h2>\n

By \"\M{n}\" we denote the set of all \"n \times n\" split-complex matrices<\/em> i.e. matrices in which entries are split-complex numbers. Note that \"Z \in \M{n}\" if and only if there exist \"n \times n\" real matrices \"A\", \"B\" such that \"Z = A + jB\". If \"A\" is a matrix then its transpose<\/em> is defined as \"a_{i,j}^T = a_{j,i}\" for all \"i,j\" and is denoted with \"A^T\". In the following we denote by \"I_n\" the \"n \times n\" identity matrix<\/em> and by \"O_n\" the \"n \times n\" zero matrix<\/em>. Let \"J \in M_n(\mathbb{R})\" be defined as \"J = \mathrm{antidiag}(1,\dots,1)\" for each \"n\". Note that \"J^2 = I\". A matrix \"R \in M_n(\mathbb{R})\" is upper-triangular<\/em> if \"r_{i,j} = 0\" for all \"i > j\".
A real matrix \"R\" is centrosymmetric<\/em> if \"RJ = JR\". An overview of centrosymmetric matrices can be found in [4]. We denote by \"\CsMat{n}\" the set of all \"n \times n\" centrosymmetric real matrices.<\/p>\n

For the algebra \"\Cs\" of split-complex numbers the well-known matrix representation \"{cs} : \Cs \rightarrow M_{2}(\mathbb{R})\" with respect to \"(e_1, e_2)\" is given by
<\/a>

(2) <\/span>   <\/span>\"\begin{equation*}{cs}(z) = \begin{bmatrix} \Real(z) & \Hyp(z) \\ \Hyp(z) & \Real(z) \end{bmatrix},\qquad \forall z \in \Cs.\end{equation*}\"<\/p>
It is straightforward to check that for all \"w, z \in \Cs\" we have \"\CsEmb{w \cdot z} = \CsEmb{w} \CsEmb{z}\".<\/p>\n

Further, on the vector space \"M_n(\Cs)\" there is a natural multiplication operation \"\cdot : M_n(\Cs) \times M_n(\Cs) \rightarrow M_n(\Cs)\" given by
<\/a>

(3) <\/span>   <\/span>\"\begin{equation*}(A + jB) \cdot (C + jD) := (AC + BD) + j(AD + BC)\end{equation*}\"<\/p>
for all \"A,B,C,D \in M_n(\mathbb{R})\". It is easy to verify that
\"(M_n(\Cs), \cdot)\" is an algebra. In the following we refer to this algebra as the algebra of split-complex (square) matrices<\/em> and denote it with \"\mathfrak{M}_n\".<\/p>\n

Note that in the following whenever we have two matrices \"A,B \in M_n(\Cs)\", their product shall explicitly be written with a dot '\"\cdot\"', e.g. \"A \cdot B\" to indicate multiplication defined in (3). Otherwise, if \"A,B \in M_n(\mathbb{R})\" we simply write \"AB\".<\/p>\n

To state and prove our main result, we shall need the following well known characterization of centrosymmetric matrices.<\/p>\n

Proposition 1<\/strong>:
Let \"A \in M_{2n}(\mathbb{R})\". Then \"A \in \CsMat{2n}\" if and only if there exist \"B,C \in M_{n}(\mathbb{R})\" such that
<\/a>

(4) <\/span>   <\/span>\"\begin{equation*}A = \begin{bmatrix}B & C J \\JC & J B J\end{bmatrix}.\end{equation*}\"<\/p><\/p>\n

Proof:<\/em>
Suppose

  <\/span>   <\/span>\"\[A =\begin{bmatrix}X & Y \\W & Z\end{bmatrix}.\]\"<\/p>
Since \"A\" is centrosymmetric, we have \"AJ = JA\", or equivalently, in block-form

*** QuickLaTeX cannot compile formula:\n\[\begin{bmatrix}\ & J \\J & \\end{bmatrix}\begin{bmatrix}X & Y \\W & Z\end{bmatrix} =\begin{bmatrix}X & Y \\W & Z\end{bmatrix}\begin{bmatrix}\ & J \\J & \\end{bmatrix}\]\n\n*** Error message:\n\begin{bmatrix} on input line 17 ended by \end{equation*}.\r\nleading text: ...begin{bmatrix}\ & J \\J & \\end{bmatrix}\]\r\nMissing $ inserted.\r\nleading text: ...begin{bmatrix}\ & J \\J & \\end{bmatrix}\]\r\nMissing } inserted.\r\nleading text: ...begin{bmatrix}\ & J \\J & \\end{bmatrix}\]\r\nMissing \cr inserted.\r\nleading text: ...begin{bmatrix}\ & J \\J & \\end{bmatrix}\]\r\nMissing $ inserted.\r\nleading text: ...begin{bmatrix}\ & J \\J & \\end{bmatrix}\]\r\nMissing } inserted.\r\nleading text: ...begin{bmatrix}\ & J \\J & \\end{bmatrix}\]\r\nMissing \right. inserted.\r\nleading text: ...begin{bmatrix}\ & J \\J & \\end{bmatrix}\]\r\nImproper \prevdepth.\r\nleading text: \end{document}\r\nMissing $ inserted.\r\nleading text: \end{document}\r\nMissing } inserted.\r\nleading text: \end{document}\r\n\n<\/pre>
This is equivalent to
   <\/span>   <\/span>\"\[\begin{bmatrix}JW & JZ \\JX & JY\\end{bmatrix} =\begin{bmatrix}YJ & XJ \\ZJ & WJ\end{bmatrix}\]\"<\/p>
We now have \"Z = JXJ\" and \"Y = JWJ\", so
   <\/span>   <\/span>\"\[A = \begin{bmatrix}X & JWJ \\W & JXJ\end{bmatrix}\]\"<\/p>
Now, by choosing \"C = JW\" and \"B = X\" and from the fact \"J^2 = I\" it follows that \"A\" has the form (4).<\/p>\n
\u00a0<\/p>\n
Conversely, suppose \"A\" has the form (4). It can easily be shown by block-matrix multiplication that \"AJ = JA\", hence \"A\" is centrosymmetric.<\/p>\n
QED<\/em>.<\/p>\n
The map \"{cs} : \M{n} \rightarrow CM_{2n}(\mathbb{R})\" defined as
<\/a>
 (5) <\/span>   <\/span>\"\begin{equation*}{cs}(A + jB) = \begin{bmatrix}A & BJ \\JB & JAJ\end{bmatrix}\end{equation*}\"<\/p>
is a matrix representation of \"\mathfrak{M}_n\". We call the representation \"{cs}\" the standard centrosymmetric matrix representation<\/em> of \"\mathfrak{M}_n\".<\/p>\n
Proof<\/em>:
Let \"W \in \M{n}\" and \"Z \in \M{n}\" be such that \"W = A + j B\" and \"Z = C + j D\".
We now have
   <\/span>   <\/span>\"\begin{align*}\CsEmb{W} \CsEmb{Z} &= \begin{bmatrix} A & B J \\ J B & J A J \end{bmatrix} \begin{bmatrix} C & D J \\ J D & J C J \end{bmatrix} \\ &= \begin{bmatrix} A C + B D & (A D + B C) J \\ J (A D + B C) & J (A C + B D) J\end{bmatrix} = \CsEmb{W \cdot Z}\end{align*}\"<\/p>
which proves the claim.<\/p>\n
QED<\/em>.<\/p>\n
Proposition<\/strong>.
Let \"Q \in \M{n}\". Then \"{cs}(Q^T) = {cs}(Q)^T\".<\/p>\n
Proof<\/em>.
Let \"Q = A + jB\". Then
   <\/span>   <\/span>\"\[{cs}(Q^T) = {cs}(A^T + jB^T) = \begin{bmatrix}A^T & B^T J \\J B^T & J A^T J\end{bmatrix}.\]\"<\/p>
On the other hand, keeping in mind that \"J^T = J\" we have
   <\/span>   <\/span>\"\[{cs}(Q)^T = \begin{bmatrix}A & B J \\J B & JAJ\end{bmatrix}^T =\begin{bmatrix}A^T & (JB)^T \\(BJ)^T & (JAJ)^T\end{bmatrix}= \begin{bmatrix}A^T & B^T J \\J B^T & J A^T J\end{bmatrix}.\]\"<\/p>
Hence, \"{cs}(Q^T) = {cs}(Q)^T\".<\/p>\n
QED<\/em>.<\/p>\n
Proposition<\/strong>.
The map \"{cs}\" is a bijection.<\/p>\n
Proof<\/em>.
Injectivity<\/em>. Let \"Z = A + jB\" and \"W = C + jD\" and \"W \not = Z\". From this,
it follows that \"A \not = C\" or \"B \not = D\". Assume that \"A \not = C\". Then
   <\/span>   <\/span>\"\begin{align*}\CsEmb{Z} = \begin{bmatrix}A & BJ \\JB & JAJ\end{bmatrix} & \quad &\CsEmb{W} = \begin{bmatrix}C & DJ \\JD & JCJ\end{bmatrix}.\end{align*}\"<\/p>
Since \"A \not = C\" we have \"\CsEmb{Z} \not = \CsEmb{W}\". Let now \"B \not = D\" and assume that \"\CsEmb{Z} = \CsEmb{W}\". Then from \"\CsEmb{Z} = \CsEmb{W}\" it follows \"JB = JD\". Now multiplying \"JB = JD\" with \"J\" from the left implies \"B=D\", which is a contradiction.
We conclude that \"\CsEmb{\cdot}\" is injective.<\/p>\n
Surjectivity<\/em>. Let \"A \in \CsMat{2n}\". By proposition 1 we can find matrices \"B\" and \"C\" such that (4) holds. But then \"\CsEmb{B + jC} = A\" and since \"A\" was arbitrary, we conclude that \"{cs}\" is surjective.
Now, injectivity and surjectivity of \"{cs}\" imply by definition that \"{cs}\" is a bijection.<\/p>\n
QED<\/em>.<\/p>\n
Correspondence of QR decompositions<\/h2>\n
Definition:<\/strong>
Let \"A \in M_n(\Cs)\". A pair \"(Q, R)\" with \"Q, R \in M_n(\Cs)\" is a QR decomposition of \"A\" over \"\Cs\"<\/em>\u00a0if the following holds:<\/p>\n
    \n
  1. \"Q\" is orthogonal, i.e. \"Q^T \cdot Q = Q \cdot Q^T = I\",<\/li>\n
  2. \"R\" is upper-triangular,<\/li>\n
  3. \"A = Q \cdot R.\"<\/li>\n<\/ol>\n
    The notion of a \"m\times n\" double-cone matrix<\/em> was introduced by the author in [1].
    Here we state the definition in block-form for the case of \"H \in CM_{2n}(\mathbb{R})\".<\/p>\n
    Definition<\/strong>:
    Let \"H \in CM_{2n}(\mathbb{R})\". Then \"H\" is a double-cone matrix<\/em> iff there exist \"A, B \in M_n(\mathbb{R})\" both upper-triangular such that
       <\/span>   <\/span>\"\[H = \begin{bmatrix}A & BJ \\JB & JAJ\end{bmatrix}.\]\"<\/p><\/p>\n
    Definition<\/strong>:
    Let \"Z \in \CsMat{n}\". A pair \"(W, Y)\", with \"W, Y \in \CsMat{n}\" is a centrosymmetric QR decomposition<\/em> of \"Z\" if the following holds:<\/p>\n
      \n
    1. \"W\" is orthogonal matrix,<\/li>\n
    2. \"Y\" is double-cone matrix,<\/li>\n
    3. \"Z = WY.\"<\/li>\n<\/ol>\n
      The algorithm to obtain an approximation of a centrosymmetric QR decomposition of a given centrosymmetric matrix \"A\" was given in [1].<\/p>\n
      The following theorem provides one interpretation of the centrosymmetric QR decomposition, in the case of square centrosymmetric matrices of even order by establishing the equivalence of their centrosymmetric QR decomposition with the QR decomposition of the corresponding split-complex matrix.<\/p>\n
      Theorem 1<\/strong> (QR decomposition correspondence)<\/em>:
      Let \"A \in \M{n}\". Then \"(Q,R) \in \M{n} \times \M{n}\" is a QR decomposition of \"A\" if and only if 
         <\/span>   <\/span>\"\[(\CsEmb{Q}, \CsEmb{R}) \in \CsMat{2n} \times \CsMat{2n}\]\"<\/p> is a centrosymmetric QR decomposition of \"\CsEmb{A} \in \CsMat{2n}\".<\/p>\n
      Proof<\/em>.
      Let \"(Q, R) \in \M{n} \times \M{n}\" be a QR decomposition of \"A\".
      Let \"W = \CsEmb{Q}\" and \"Y = \CsEmb{R}\". We have
         <\/span>   <\/span>\"\begin{align*}W = \begin{bmatrix}Q_1 & Q_2 J \\J Q_2 & J Q_1 J\end{bmatrix} & \ &Y = \begin{bmatrix}R_1 & R_2 J \\J R_2 & J R_1 J\end{bmatrix}\end{align*}\"<\/p>
      Since \"Q^T \cdot Q = I\" it follows that \"\CsEmb{Q^T \cdot Q} = \CsEmb{Q^T} \CsEmb{Q} = \CsEmb{Q}^T \CsEmb{Q} = \CsEmb{I}\".
      From this we have \"\CsEmb{Q}^T \CsEmb{Q} = I\" i.e. \"W^T W = I\" hence \"W\" is orthogonal. Since \"R\" is upper-triangular and \"R = R_1 + j R_2\", then by definition we have that both \"R_1\" and \"R_2\" are upper-triangular. Further, \"\CsEmb{R}\" is centrosymmetric by definition. From this it follows that \"Y\" is centrosymmetric double-cone.
      Finally, we have \"\CsEmb{A} = \CsEmb{Q \cdot R} = \CsEmb{Q} \CsEmb{R}\". Hence, \"(\CsEmb{Q}, \CsEmb{R})\" is
      a centrosymmetric QR decomposition of \"\CsEmb{A}\".<\/p>\n
      Conversely, let \"(W, Y) = (\CsEmb{Q}, \CsEmb{R})\". If \"(W, Y)\" is a centrosymmetric QR decomposition of \"\CsEmb{A}\" then \"\CsEmb{A} = WY\" where \"W\" is centrosymmetric and orthogonal and \"Y\" is a double-cone matrix.<\/p>\n
      From the fact that \"W\" is centrosymmetric we have (by Proposition 1) that
         <\/span>   <\/span>\"\[W = \begin{bmatrix}W_1 & W_2 J \\J W_2 & J W_1 J\end{bmatrix}\]\"<\/p>
      Now the property of \"W\" being orthogonal i.e. the condition \"W^T W = I\" implies
      <\/a>
       (6) <\/span>   <\/span>\"\begin{equation*}\begin{bmatrix}W_1^T W_1 + W_2^T W_2 & (W_1^T W_2 + W_2^T W_1) J \\J(W_2^T W_1 + W_1^T W_2) & J(W_2^T W_2 + W_1^T W_1) J\end{bmatrix}= \begin{bmatrix}I & \ \\\ & I\end{bmatrix}.\end{equation*}\"<\/p>
      On the other hand, we have
      <\/a>
       (7) <\/span>   <\/span>\"\begin{align*}Q = W_1 + j W_2 & \ & R = Y_1 + j Y_2\end{align*}\"<\/p>
      First we prove that \"Q\" is orthogonal. From (6) we obtain
         <\/span>   <\/span>\"\begin{align*}Q^T \cdot Q &= (W_1 + j W_2)^T \cdot (W_1 + j W_2) \\&= (W_1^T + j W_2^T) \cdot (W_1 + j W_2) \\& = W_1^T W_1 + W_2^T W_2 + j (W_1^T W_2 + W_2^T W_1) \\& = I + jO = I.\end{align*}\"<\/p>
      The matrix \"Y\" is centrosymmetric and double-cone which implies
         <\/span>   <\/span>\"\[Y = \begin{bmatrix}Y_1 & Y_2 J \\J Y_2 & J Y_1 J\end{bmatrix}\]\"<\/p>
      where both \"Y_1\" and \"Y_2\" are upper-triangular. This now implies that \"{cs}^{-1}(Y) = Y_1 + j Y_2\" is upper-triangular.<\/p>\n
      Finally, let us prove that \"QR = A\". We have
         <\/span>   <\/span>\"\[Q \cdot R = {cs}^{-1} (W) {cs}^{-1} (Y) = {cs}^{-1}(WY) = {cs}^{-1}(\CsEmb{A}) = A.\]\"<\/p>
      We conclude that \"(Q, R)\" is a QR decomposition of \"A\".
      QED<\/em>.<\/p>\n
      Example:<\/em>
      Let
         <\/span>   <\/span>\"\[A = \begin{bmatrix}1 + 2j & 2 + 3j \\3 + 4j & 4 + 5j\end{bmatrix}.\]\"<\/p>
      Note that \"A = W + jZ\" where
       (8) <\/span>   <\/span>\"\begin{align*}W = \begin{bmatrix}1 & 2 \\3 & 4\end{bmatrix} & \mbox{ and } Z = \begin{bmatrix}2 & 3 \\4 & 5\end{bmatrix}.\end{align*}\"<\/p>
      We have
         <\/span>   <\/span>\"\[{cs}(A) = {cs}(W + jZ) = \begin{bmatrix}W & ZJ \\JZ & JWJ\end{bmatrix}= \begin{bmatrix}1 & 2 & 3 & 2 \\3 & 4 & 5 & 4 \\4 & 5 & 4 & 3 \\2 & 3 & 2 & 1\end{bmatrix}\]\"<\/p>
      By applying the \"\textsc{CentrosymmetricQR}\" algorithm from [1] to \"{cs}(A)\" we obtain the approximations:<\/p>\n
         <\/span>   <\/span>\"\begin{align*}Q \approx \begin{bmatrix}-0.156594 & -0.106019 & -0.813126 & 0.550513 \\0.106019 & -0.156594 & 0.550513 & 0.813126 \\0.813126 & 0.550513 & -0.156594 & 0.106019 \\0.550513 & -0.813126 & -0.106019 & -0.156594 \\\end{bmatrix}\\\\\R \approx \begin{bmatrix}4.51499 & 5.82806 & 4.41384 & 3.10078 \\0 & -0.525226 & -0.525226 & 0 \\0 & -0.525226 & -0.525226 & 0 \\3.10078 & 4.41384 & 5.82806 & 4.51499 \\\end{bmatrix}\end{align*}\"<\/p><\/p>\n
      Applying \"{cs}^{-1}\" to \"Q\" and \"R\" yields:
         <\/span>   <\/span>\"\begin{align*}{cs}^{-1}(Q) \approx \begin{bmatrix}-0.156594 & -0.106019 \\0.106019 & -0.156594\end{bmatrix} + j \begin{bmatrix}0.550513 & -0.813126 \\0.813126 & 0.550513\end{bmatrix} \\& \ & \\{cs}^{-1}(R) \approx \begin{bmatrix}4.51499 & 5.82806 \\0 & -0.525226\end{bmatrix} + j\begin{bmatrix}3.10078 & 4.41384 \\0 & -0.525226 \\\end{bmatrix}\end{align*}\"<\/p>
      with \"A \approx {cs}^{-1}(Q) \cdot {cs}^{-1}(R)\". Now, from Theorem 1 we conclude that \"({cs}^{-1}(Q), {cs}^{-1}(R))\" is an approximation of a QR decomposition of \"A\".<\/p>\n
      Conclusion<\/h2>\n
      We introduced the standard centrosymmetric representation \"{cs}\" for split-complex matrices. Using this representation we proved that a QR decomposition of a square split-complex matrix \"A\" can be obtained by calculating the centrosymmetric QR decomposition introduced by the author in [1] of its centrosymmetric matrix representation \"{cs}(A)\".<\/p>\n
      References<\/h2>\n
        \n
      1. \u00a0<\/span>Burnik,\u00a0Konrad<\/span><\/span><\/span>.\u00a0<\/span><\/em>A\u00a0structure-preserving\u00a0<\/span><\/span>QR f<\/span>actorization\u00a0for\u00a0cen<\/span><\/span>trosymmetric\u00a0real\u00a0matrices,
        <\/span><\/span><\/em>Linear\u00a0Algebra\u00a0and\u00a0its\u00a0Applications\u00a0484<\/span><\/span><\/span><\/span>(2015) 356 - 378\u00a0<\/span><\/em><\/li>\n
      2. Catoni,\u00a0Francesco\u00a0and\u00a0Boccaletti,\u00a0Dino\u00a0and\u00a0Cannata,\u00a0<\/span><\/span><\/span><\/span>Roberto\u00a0and\u00a0Catoni,\u00a0Vincenzo\u00a0and\u00a0Zampeti,\u00a0Paolo<\/span><\/span><\/span><\/span><\/span>, Hyper<\/span>bolic Numbers.\u00a0<\/span><\/em>Geometry of Minkowski Space-Time Springer Berlin Hei<\/span>delberg\u00a0Berlin,\u00a0Heidelberg,\u00a0(2011)\u00a03\u201323\u00a0ISBN: 978-3-642-17977-8<\/span><\/span><\/em><\/li>\n
      3. Curtis, Charles W.; Reiner, Irving.\u00a0Representation Theory of Fi<\/span>nite Groups and Associative Algebras,\u00a0<\/span><\/em>John Wiley & Sons (Reedition<\/span>2006\u00a0by AMS\u00a0Bookstore), (1962)\u00a0ISBN 978-0-470-18975<\/span><\/span><\/em><\/li>\n
      4. James\u00a0R.\u00a0Wea<\/span><\/span><\/span>ver.\u00a0<\/span><\/span><\/span><\/span><\/span>Centrosymmetric\u00a0(Cross-Symmetric)\u00a0Matrices,<\/span><\/span>Their\u00a0Basic Properties,\u00a0Eigenvalues,\u00a0and\u00a0Eigenvectors,\u00a0<\/span><\/span><\/span><\/span><\/span><\/em>The\u00a0American\u00a0<\/span><\/span><\/span><\/span><\/span>Mathematical Monthly, Mathematical Association of America, (1985)\u00a0<\/span><\/em>10-<\/span>92,\u00a0711\u2013717 ISSN:\u00a000029890,\u00a019300972\u00a0doi:10.2307\/2323222<\/span><\/span><\/em><\/li>\n<\/ol>\n\nQR_decomposition_of_split_complex_matric<\/a>\n\n\n
        Copyright \u00a9 2018, Konrad Burnik<\/p>\n
        0<\/span><\/a><\/div> <\/div> 
        <\/div><\/div>
        <\/div>","protected":false},"excerpt":{"rendered":"
        Introduction Back in 2015 I published an article [1] about the QR decomposition of a centrosymmetric real matrix. In 2016 I started thinking about the meaning of this decomposition and centrosymmetric matrices in general. I discovered that centrosymmetric matrices have a natural interpretation as split-complex matrices and even more, that the centrosymmetric QR decomposition of […]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"_mi_skip_tracking":false,"_exactmetrics_sitenote_active":false,"_exactmetrics_sitenote_note":"","_exactmetrics_sitenote_category":0,"_s2mail":"yes","footnotes":""},"categories":[7],"tags":[],"class_list":["post-652","post","type-post","status-publish","format-standard","hentry","category-matrix-decompositions"],"_links":{"self":[{"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/posts\/652","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/comments?post=652"}],"version-history":[{"count":62,"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/posts\/652\/revisions"}],"predecessor-version":[{"id":1302,"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/posts\/652\/revisions\/1302"}],"wp:attachment":[{"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/media?parent=652"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/categories?post=652"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/konrad.burnik.org\/wordpress\/wp-json\/wp\/v2\/tags?post=652"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}