{"id":88,"date":"2009-04-25T17:54:34","date_gmt":"2009-04-25T15:54:34","guid":{"rendered":"http:\/\/wp1.fredptitgars.net\/index.php\/2009\/04\/25\/formule-latex\/"},"modified":"2009-04-25T17:54:34","modified_gmt":"2009-04-25T15:54:34","slug":"formule-latex","status":"publish","type":"post","link":"https:\/\/fredptitgars.ovh\/?p=88","title":{"rendered":"formule latex"},"content":{"rendered":"<p><math><\/p>\n<p>La factorisation de Cholesky, consiste, pour une matrice sym\u00e9trique d\u00e9finie positive $A$, \u00e0 d\u00e9terminer une matrice triangulaire inf\u00e9rieure $L$ tel que $A=LL^T$. Une matrice sym\u00e9trique $A$ est dite d\u00e9finie positive si, pour tout vecteur $x$, le produit $x^TAx$ est positif.<\/p>\n<p>La matrice $L$ est en quelque sorte une \u00ab racine carr\u00e9e \u00bb de $A$. Cette d\u00e9composition permet notamment de calculer la matrice inverse $A^<em>-1<\/em>$, de calculer le d\u00e9terminant de $A$ (\u00e9gal au carr\u00e9 du produit des \u00e9l\u00e9ments diagonaux\nde $L$).<\/p>\n<h2>Exemple<\/h2>\n<p>La matrice sym\u00e9trique\n$[C_3] = \n<em>\\color<em>red<\/em> \\langle \\vec k_<em>3L<\/em> \\vec k_<em>3L<\/em>^<em>\\:\\dagger<\/em> \\rangle <\/em>=\n\\left[ \\begin<em>array<\/em><em>rrrr<\/em>\nS_<em>11<\/em> &#038; S_<em>21<\/em> &#038; S_<em>31<\/em> \\\\\nS_<em>12<\/em> &#038; S_<em>22<\/em> &#038; S_<em>32<\/em> \\\\\nS_<em>13<\/em> &#038; S_<em>23<\/em> &#038; S_<em>33<\/em> \n\\end<em>array<\/em> \\right] $<\/p>\n<p>est \u00e9gale au produit de la matrice triangulaire $L$ et de sa transpos\u00e9e $L^T$:<\/p>\n<p>$$\n\\pmatrix<em>\n1 &#038; 1 &#038; 1  &#038; 1 \\cr\n1 &#038; 5 &#038; 5  &#038; 5 \\cr\n1 &#038; 5 &#038; 14 &#038; 14 \\cr\n1 &#038; 5 &#038; 14 &#038; 15\n<\/em>=\n\\pmatrix<em>\n1 &#038; 0 &#038; 0  &#038; 0 \\cr\n1 &#038; 2 &#038; 0  &#038; 0 \\cr\n1 &#038; 2 &#038; 3  &#038; 0 \\cr\n1 &#038; 2 &#038; 3  &#038; 1\n<\/em>\\dot\n\\pmatrix<em>\n1 &#038; 1 &#038; 1  &#038; 1 \\cr\n0 &#038; 2 &#038; 2  &#038; 2 \\cr\n0 &#038; 0 &#038; 3  &#038; 3 \\cr\n0 &#038; 0 &#038; 0  &#038; 1\n<\/em>$$<\/p>\n<p>avec<\/p>\n<p>$$L=\n\\pmatrix<em>\n1 &#038; 0 &#038; 0  &#038; 0 \\cr\n1 &#038; 2 &#038; 0  &#038; 0 \\cr\n1 &#038; 2 &#038; 3  &#038; 0 \\cr\n1 &#038; 2 &#038; 3  &#038; 1\n<\/em>$$<\/p>\n<h2>Th\u00e9or\u00e8me<\/h2>\n<p>Factorisation de Cholesky d&rsquo;une matrice :<\/p>\n<p>Si $A$ est une matrice sym\u00e9trique d\u00e9finie positive, il existe au\nmoins une matrice r\u00e9elle triangulaire inf\u00e9rieure $L$ telle que :\n$$A=LL^T$$<\/p>\n<p>On peut \u00e9galement imposer que les \u00e9l\u00e9ments diagonaux de la matrice\n$L$ soient tous positifs, et la factorisation correspondante\nest alors unique.<\/p>\n<h2>Algorithme<\/h2>\n<p>On cherche la matrice :<\/p>\n<p>$$\nL=\\pmatrix<em>\nl_<em>11<\/em>\\cr\nl_<em>21<\/em> &#038; l_<em>22<\/em>\\cr\n\\vdots &#038; \\vdots &#038; \\ddots\\cr\nl_<em>n1<\/em> &#038; l_<em>n2<\/em> &#038; \\cdots &#038; l_<em>nn<\/em>\n<\/em>\n$$<\/p>\n<p>De l&rsquo;\u00e9galit\u00e9 $A=LL^T$ on d\u00e9duit :\n$a_<em>ij<\/em>=\\left(LL^<em>T<\/em>\\right)_<em>ij<\/em>=<em>\\sum_<em>k=1<\/em>^<em>n<\/em>l_<em>ik<\/em>l_<em>jk<\/strong>=<em>\\sum_<em>k=1<\/em>^<em>\\min\\left\\<em> i,j\\right\\<\/em> <\/em>l_<em>ik<\/em>l_<em>jk<\/strong>,\\;1\\leq i,j\\leq n$<\/p>\n<p>puisque $l_<em>ij<\/em>=0$ si $1 \\leq i < j \\leq n.$\n\nLa matrice $A$ \u00e9tant sym\u00e9trique, il suffit que les relations ci-dessus soient v\u00e9rifi\u00e9es pour i\u2264j, c'est-\u00e0-dire que les \u00e9l\u00e9ments l<sub>ij<\/sub> de la matrice $L$ doivent satisfaire :\n$a_<em>ij<\/em>=<em>\\sum_<em>k=1<\/em>^<em>i<\/em>l_<em>ik<\/em>l_<em>jk<\/strong>,\\;1\\leq i,j\\leq n$<\/p>\n<p>Pour j=1, on d\u00e9termine la premi\u00e8re colonne de $L$ :\n&#8211; (i=1) $a_<em>11<\/em>=l_<em>11<\/em>l_<em>11<\/em>$ d&rsquo;o\u00f9 $l_<em>11<\/em>=\\sqrt<em>a_<em>11<\/strong>$\n&#8211; (i=2) $a_<em>12<\/em>=l_<em>11<\/em>l_<em>21<\/em>$ d&rsquo;o\u00f9 $l_<em>21<\/em>=\\frac<em>a_<em>12<\/strong><em>l_<em>11<\/strong>$\n&#8211; &#8230;\n&#8211; (i=n) $a_<em>1n<\/em>=l_<em>11<\/em>l_<em>n1<\/em>$ d&rsquo;o\u00f9 $l_<em>n1<\/em>=\\frac<em>a_<em>1n<\/strong><em>l_<em>11<\/strong>$<\/p>\n<p>On d\u00e9termine la j-\u00e8me colonne de $L$, apr\u00e8s avoir calcul\u00e9 les (j-1) premi\u00e8res colonnes :\n&#8211; (i=j) $a_<em>ii<\/em>=l_<em>i1<\/em>l_<em>i1<\/em>+\\ldots+l_<em>ii<\/em>l_<em>ii<\/em>$ d&rsquo;o\u00f9 $l_<em>ii<\/em>=\\sqrt<em>a_<em>ii<\/em>&#8211;<em>\\sum_<em>k=1<\/em>^<em>i-1<\/em>l_<em>ik<\/em>^<em>2<\/strong><\/em>$\n&#8211; (i=j+1) $\\displaystyle a_<em>i,i+1<\/em>=l_<em>i1<\/em>l_<em>i+1<\/em>+\\ldots+l_<em>ii<\/em>l_<em>i+1,i<\/em>$ d&rsquo;o\u00f9 $l_<em>i+1,i<\/em>=\\frac<em>a_<em>i,i+1<\/em>&#8211;<em>\\sum_<em>k=1<\/em>^<em>i-1<\/em>l_<em>ik<\/em>l_<em>i+1,k<\/strong><\/em><em>l_<em>ii<\/strong>$\n&#8211; &#8230;\n&#8211; (i=n) $\\displaystyle a_<em>i,n<\/em>=l_<em>i1<\/em>l_<em>n1<\/em>+\\ldots+l_<em>ii<\/em>l_<em>ni<\/em>$ d&rsquo;o\u00f9 $l_<em>ni<\/em>=\\frac<em>a_<em>in<\/em>&#8211;<em>\\sum_<em>k=1<\/em>^<em>i-1<\/em>l_<em>ik<\/em>l_<em>nk<\/strong><\/em><em>l_<em>ii<\/strong>$<\/p>\n<h2>R\u00e9solution de syst\u00e8me<\/h2>\n<p>Pour la r\u00e9solution de syst\u00e8me lin\u00e9aire de la forme:$Ax=b$, le syst\u00e8me devient<\/p>\n<p>$$LL^Tx = b \\Leftrightarrow\n\\left\\<em>\\begin<em>array<\/em><em>cc<\/em>\nLy = b&#038; (1),\\\\\nL^Tx = y &#038;(2).\n\\end<em>array<\/em>\\right.\n$$<\/p>\n<p>On r\u00e9sout le syst\u00e8me (1) pour trouver le vecteur $y$, puis le syst\u00e8me (2) pour trouver le vecteur $x$. La r\u00e9solution est facilit\u00e9e par la forme triangulaire des matrices.<\/p>\n<h2>Calcul de d\u00e9terminant<\/h2>\n<p>La m\u00e9thode de Cholesky permet aussi de calculer le d\u00e9terminant de A, qui est \u00e9gal au carr\u00e9 du produit des \u00e9l\u00e9ments diagonaux de la matrice L, puisque<\/p>\n<p>$$det(A) = det(L) \u00d7 det(L^T)=det(L)^2$$<\/p>\n<p><\/math><\/p>\n","protected":false},"excerpt":{"rendered":"<p>La factorisation de Cholesky, consiste, pour une matrice sym\u00e9trique d\u00e9finie positive $A$, \u00e0 d\u00e9terminer une matrice triangulaire inf\u00e9rieure $L$ tel que $A=LL^T$. Une matrice sym\u00e9trique $A$ est dite d\u00e9finie positive si, pour tout vecteur $x$, le produit $x^TAx$ est positif. La matrice $L$ est en quelque sorte une \u00ab racine carr\u00e9e \u00bb de $A$. Cette [&hellip;]<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"closed","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[15],"tags":[],"class_list":["post-88","post","type-post","status-publish","format-standard","hentry","category-spip"],"blocksy_meta":[],"_links":{"self":[{"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=\/wp\/v2\/posts\/88","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=88"}],"version-history":[{"count":0,"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=\/wp\/v2\/posts\/88\/revisions"}],"wp:attachment":[{"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=88"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=88"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/fredptitgars.ovh\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=88"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}