nvidia-AI.summerschool17-no.../teaching-lw/205-quadratics.ipynb

726 lines
511 KiB
Text
Raw Permalink Normal View History

2017-09-22 17:26:46 +00:00
{
"metadata": {
"name": "205-quadratics"
},
"nbformat": 3,
"nbformat_minor": 0,
"worksheets": [
{
"cells": [
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Title: Symmetric Linear Systems and Quadratic Forms\n",
"Author: Thomas Breuel\n",
"Institution: UniKL"
]
},
{
"cell_type": "heading",
"level": 1,
"metadata": {},
"source": [
"Symmetric Linear Systems"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(system of linear equations)\n",
"\n",
"\\begin{eqnarray}\n",
"2 x_1 + 4 x_2 & = & 2 \\\\\\\n",
"4 x_1 + 11 x_2 & = & 1 \\\\\\\n",
"\\end{eqnarray}"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(system of linear equations in matrix form)\n",
"\n",
"$$ M x = b $$\n",
"\n",
"$$ M = \\left(\\begin{array}{cc}2 & 4 \\\\\\ 4 & 11 \\\\\\ \\end{array}\\right) $$\n",
"\n",
"Note that $M$ is symmetric in this case. Symmetric linear systems are special."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(factorizations)\n",
"\n",
"There are a number of common factorizations of the matrix $M$:\n",
"\n",
"- $A^TCA$ - source of problem, dynamical systems or statistics; $A$ may be a rectangular matrix\n",
"- $LDL^T$ - Gaussian elimination, $L$: lower triangular, $D$: diagonal\n",
"- $Q\\Lambda Q^T$ - orthogonal $Q$ (eigenvectors), diagonal $\\Lambda$ (eigenvalues)"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(minimization vs solution)\n",
"\n",
"minimize $\\frac{1}{2} x^T M x - x^T b$\n",
"\n",
"solve $M x = b$\n",
"\n",
"Note that you can derive this basically by differentiation:\n",
"\n",
"$$\\nabla_x (\\frac{1}{2} x^T M x - x^T b) = M x - b = 0$$"
]
},
{
"cell_type": "heading",
"level": 1,
"metadata": {},
"source": [
"Quadratic Forms"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(quadratic forms)\n",
"\n",
"Quadratic forms are \"quadratic functions of vectors\", like this:\n",
"\n",
"$$ f(x) = u x_1^2 + v x_2 ^ 2 $$\n",
"\n",
"We write a quadratic as with a matrix, like this:\n",
"\n",
"$$ f(x) = x^T M x $$\n",
"\n",
"where\n",
"\n",
"$$ M = \\left(\\begin{array}{cc}u & 0 \\\\\\ 0 & v \\\\\\ \\end{array}\\right) $$\n",
"\n",
"Therefore, a matrix $M$ is sometimes used as a linear function, $f(x) = Mx$, and sometimes as a quadratic form $f(x) = x^TMx$."
]
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# positive definite\n",
"from mpl_toolkits.mplot3d import Axes3D\n",
"subplots(1,1,figsize=(8,8))\n",
"x,y = meshgrid(linspace(-1.2,1.2,20),linspace(-1.2,1.2,20))\n",
"\n",
"z = x**2+y**2\n",
"\n",
"ax = gcf().add_subplot(111,projection='3d')\n",
"ax.plot_wireframe(x,y,z)\n",
"ax.set_zlim3d(-3.0,3.0)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 3,
"text": [
"(-3.0, 3.0)"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAcwAAAHMCAYAAABY25iGAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nMcLevTBp0oE11aQu4wioc434
}
],
"prompt_number": 3
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# indefinite\n",
"z = x**2-y**2\n",
"\n",
"ax = gcf().add_subplot(111,projection='3d')\n",
"ax.plot_wireframe(x,y,z)\n",
"ax.set_zlim3d(-3.0,3.0)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 4,
"text": [
"(-3.0, 3.0)"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAV0AAADtCAYAAAAcNaZ2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nrNhibB97r9eb9PniWGLv3WRp
}
],
"prompt_number": 4
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# negative definite\n",
"z = -x**2-y**2\n",
"\n",
"ax = gcf().add_subplot(111,projection='3d')\n",
"ax.plot_wireframe(x,y,z)\n",
"ax.set_zlim3d(-3.0,3.0)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 5,
"text": [
"(-3.0, 3.0)"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAV0AAADtCAYAAAAcNaZ2AAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nny92wnnvFktL9Z+OdJ1k69TV
}
],
"prompt_number": 5
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(definiteness of matrices)\n",
"\n",
"positive definite:\n",
"\n",
"$$ M = \\left(\\begin{array}{cc}1 & 0 \\\\\\ 0 & 1 \\\\\\ \\end{array}\\right) $$\n",
"\n",
"indefinite:\n",
"\n",
"$$ M = \\left(\\begin{array}{cc}1 & 0 \\\\\\ 0 & -1 \\\\\\ \\end{array}\\right) $$\n",
"\n",
"negative definite:\n",
"\n",
"$$ M = \\left(\\begin{array}{cc}-1 & 0 \\\\\\ 0 & -1 \\\\\\ \\end{array}\\right) $$\n"
]
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# level sets and contour plot\n",
"figsize(5,5)\n",
"x,y = meshgrid(linspace(-1.2,1.2,20),linspace(-1.2,1.2,20))\n",
"z = x**2+y**2\n",
"contour(x,y,z,levels=[0.5,1,1.5])"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 6,
"text": [
"<matplotlib.contour.QuadContourSet instance at 0x326a950>"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAT8AAAExCAYAAAAKrKXbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nL/xBEpv5iQGE0pFOasspRQjB
}
],
"prompt_number": 6
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# ellipses\n",
"figsize(5,5)\n",
"x,y = meshgrid(linspace(-1.2,1.2,20),linspace(-1.2,1.2,20))\n",
"z = x**2+2*y**2\n",
"contour(x,y,z,levels=[0.5,1,1.5])"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 7,
"text": [
"<matplotlib.contour.QuadContourSet instance at 0x2f7f878>"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAT8AAAExCAYAAAAKrKXbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nH35Ij48/pvWIEZrWI+FXiqQD
}
],
"prompt_number": 7
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# quadratic form in intensity\n",
"imshow(z,cmap=cm.gray)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 8,
"text": [
"<matplotlib.image.AxesImage at 0x3718d10>"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAATQAAAExCAYAAADyTF4sAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\npFf0bzdRL3PuzHbLORa8ULAN
}
],
"prompt_number": 8
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# evaluation of quadratic form in matrix form\n",
"M = array([[3.0,0.0],[0.0,1.0]])\n",
"z = apply_along_axis(lambda x:dot(x,dot(M,x)),0,array([x,y]))\n",
"contour(x,y,z,levels=[0.5,1,1.5])"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 9,
"text": [
"<matplotlib.contour.QuadContourSet instance at 0x3739c68>"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAT8AAAExCAYAAAAKrKXbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz\nAAALEgAACxIB0t1+/AAAIABJREFUeJztnXd0VlX2v583hYQEUggkQBIMJZRQQkKJ2IjSW0AEREcF\nYQRRZKqj48wozNdxdEZn/XRQAcdRmFFKkKqA1MBQQwlFEnqCKRA6gUD6/f2BQcAEwr3n3rftZy3W\nInDvPvvlJh/25+xzz7FpmqYhCILgZnjYOwFBEAR7IOInCIJbIuInCIJbIuInCIJbIuInCIJbIuIn\nCIJb4mXvBCqx2Wz2TkEQBBelqhV9DlX5aZpmyq833njDtNiO9MuRPmfx5cvsmDGDWb168deAAJJH\njGB/cjIlhYV2/5wVFRWc2L2bNa+9xgfR0fy/pk1Z9bvfkf/dd3b/d3PU5+nMn7M6HKbyE1yHrPXr\nWTJmDKHt2tFp3DhGLlqEt5+fvdO6js1mo2FsLA1jY3n4zTfJ37uX9HnzmPXII8SNHUv3N97Ay8fH\n3mkKJiPiJyijpLCQNa++SsbChQz4+GNaDRpk75TuyI1C2HXSJL6ZMIEZ8fEM/vxzwrt0sXd6gok4\nlO01i8TERHunYAn2/JxZ69czrUMHigsKmLBvn6nCZ9bnrBMWxoivvuLBP/6R2QMHsua11ygrLjZl\nrJog37fmYtNuZ4otxGaz3dafC46JM1Z7NeFyfj7fTJjA2YMHpQp0cqrTFreo/ARzsLLasxpHqwIF\n9UjlJ9w116u9BQsYMG2aS4leVUgV6NxI5ScoobLaK7p40eWqveqQKtA1kcpPqBHuVu1Vh1SBzodU\nfoJu3LHaqw6pAl0HqfyEaqkoK2Plb35D+vz5bl3tVceNVeBjc+YQ1r69vVMSqqA6bRHxE6qkoqyM\nr558kuKCAh778ktq16tn75QcEk3T2DNrFqtfeYWnvv2WhrGx9k5JuIXqtEXe8BB+QqXwlVy6xMhF\ni/Dy9bV3Sg6LzWaj46hR1PL35799+ogAOhEifsJNaJrG4jFjKLl0iccXLhThqyExw4YB8EXfvoxe\nv56Qli3tnJFwJ0T8hJtY+9prnDt8mGfWrBHhu0tihg2j6OJFvujXjzGbN1MnLMzeKQm3Qbq9wnVS\nP/yQjAULeGLpUofahcWZiB87lg7PPMOXAwZQcvmyvdMRboM0PAQAMhYuZNmLLzJm40aCmzWzdzpO\njaZpfD1uHAU5OYxcsgRPb297p+TWyDo/oVq+37SJr8eN44mlS0X4FGCz2Rjw8cfYPDz4etw4+U/d\nQTEsfmPGjCEsLIz2t1njNGnSJKKjo4mNjSUtLc3okIJCTmdkMG/oUB79z39o3KmTvdNxGTy8vBg2\nbx6n9u8n5Y037J2OUAWGxe/ZZ59lxYoV1f79smXLOHLkCIcPH2bGjBlMmDDB6JCCIi6dOMEX/frR\n8513aNG3r73TcTlq+fvz5Ndfs+/LL9kxfbq90xFuwbD4PfjggwQHB1f790uWLGHUqFEAJCQkcOHC\nBfLz840OKxikuKCAL/v3J/655+g4erS903FZ/ENDeWrFCtZPmcLBpUvtnY5wA6bP+eXm5hIZGXn9\n64iICHJycsweVrgN5SUlzHvsMcITEnjwtdfsnY7LU69FC0YuXsySMWPI2brV3ukIP2DJOr9bJ3yr\nO6Zy8uTJ13+fmJjoNtt4W4mmaSwZOxZvPz/6T53qcEeGFlLIPvZykYu67vfEk1a0IoJIbDjOZwvv\n0oXBn3/O3EcflUXQJpOSkkJKSsodr1Oy1CUrK4tBgwaxb9++n/zd888/T2JiIiNHjgSgdevWrF+/\nnrBbFoDKUhdrWPP735OVksIza9Y4zFq+Mso4zCHS2EUWmbSiNWHoWyBcRDH7+Q6AOOKJpSOBBKpM\n1xC7Pv2UjW+9JYugLcRu7/YmJSUxdepURo4cydatWwkKCvqJ8AnWULmIecymTQ4hfCc4QRq72Mce\n6tOAOOIZyjB8MfZmSQ96kkM2aeziI/5JOOF0JJ42xOCNfdfcxY8dS0FODl8OGMDolBRq1alj13zc\nGcOV3xNPPMH69es5c+YMYWFhTJkyhdLSUgDGjx8PwMSJE1mxYgX+/v589tlnxMfH/zQRqfxMxVEW\nMV/mMvvYQxq7uEoRHYmjI3GEEGLKeKWUkkE6u9lFLrm0pR1xxNvVFssiaGuRLa3cmOzNm5kzZAg/\nW77cLmv5qrK1ccQTRVM8LFxnf5GL7GE3aewC7GuLK8rKmDN4MP5hYSR9+qnDzb26EiJ+bkrRhQt8\n1K4dg2bMILp/f0vHziefney4ydbG0NawrTWKhnbdFu/nO8IJJ45OtCEGLwv3+igpLOSzBx6gy8SJ\nxI8da9m47oaIn5uy/KWXKC8pYaCFi2wrqOBbVrCPPXSiMx2JN83WGqXSFu9kBwUU8BRPE0J9y8Y/\nuWcP/+nVixczMvALccx/I2dHxM8Nyd+3j1mPPHLtB6u+NT/QxRTzFcmUUMzjPEltalsyrgp2sJ21\nrOZxnuAeoiwbd9nEiaBp9P/wQ8vGdCdE/NwMTdOY1aMHbYYOpevEiZaMWUABXzCLhjRiEIMttZCq\nOMIRvmIefelPLB0tGfPquXN82KYNT61cKbtAm4Ds6uJmZCxYwJXTp+n8/POWjHeSE3zCNNrSniEM\ndUrhA2hBC55lLGtYzTrWoGH+f8i169Wj++TJrPjFL6QAsBARPxek9OpVVv7mN/R9/308vMwXoUMc\nZCaf0Zu+PER3h3qzQg+hhDGO5znEIRYwnzLKTB+z07hxFF24QPr8+aaPJVxDxM8F2fz3v9O4c2ea\nPvKI6WNtYyuLWMCTPEV7Opg+nlXUoQ7PMpZSSpnJZ1zhiqnjeXh60u+DD1j1299SesXcsYRryJyf\ni3Hx+++ZHhfHuJ07CYqKMm2cax3d5RzmME/xDPVwzaMtK6hgNSvJIJ2neMb0TvD8xx+nfps2JN7w\nnrtgDGl4uAnzR46kfqtWJE6ZYtoYztzR1YtVneCL33/P9Pj4a/953XOPaeO4EyJ+bkDW+vUsfPpp\nJh44YNq7u5Ud3UY0ZiBJTtvY0INVneD1f/4zp/btY3hysmljuBPS7XVxKsrKWDFpEr3+/nfThO/G\nju5gHnUr4QPrOsH3vfwyudu3k7lunSnxhWuI+LkIOz/5BN+gINqOGGFKfFfr6OrFik6wd+3a9H7v\nPVZMmkRFmfmdZndFxM8FuHruHOsnT6bvBx+Y8oK8q3Z09WJFJ7jN0KH4NWggZ3+YiMz5uQDLX3qJ\nirIyBnz8sdK4GhorWM5hDrl0R1cvN3aCn2a08n+f/H37mNWjh7z3axBpeLgo139A0tOVv7+7mY3s\nYQ+jGeMWHV29bGETO9jBOJ7HBx+lsZe/9BIV5eUM+OgjpXHdCWl4uCjf/upXdH/9deXC9z3H2cj/\neMJNlrIY4V7uI4IIlrJYeezEKVPI+Oor8qs4IkIwhoifE5OVksKFrCzl7++WUMIC5jOIwQRR/bGk\nwjVs2BhIErnkXj8/RBW169Xj/ldfZb0selaOiJ8TkzJ5Mg/96U/K399dzSoiiKQNMUrjujLeePMo\nQ/mGpRRSqDR25/Hjyd68mZN79iiN6+6I+DkpWSkpFOTk0OFnP1Ma9zhZ7Gcf/RmoNK470IR76EBH\nvkHt4eTefn7c97vfseHPf1Ya190R8XNSzKj6SihhIV8xkCT8sP/pbs5ID3pyghPK7a9Uf+oR8XNC\nzKr6xO4axyz7K9WfekT8nBAzqj6xu+owy/5K9acWET8nw4yqT+yuesywv1L9qUXEz8kwo+oTu6se\ns+yvVH/qEPFzIsyo+sTumocZ9leqP3WI+DkRqqs+sbvmY4b9lepPDSJ+TsLxDRuUV31id83HDPt7\nvfr7v/9TEs9dEfFzElKnTuXeX/1KWdUndtc6zLC/ncaNI3PtWgpyc5XFdDdE/JyAy/n5HFu1ig5P\nPaUknthd61Ftf2v5+9Nu5EjS
}
],
"prompt_number": 9
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# 2D rotation matrix\n",
"def rotation(a):\n",
" return array([[cos(a),-sin(a)],[sin(a),cos(a)]])\n",
"\n",
"print rotation(0.7)\n",
"print dot(rotation(0.7).T,rotation(0.7))\n",
"print dot(rotation(0.7),rotation(0.7).T)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "stream",
"stream": "stdout",
"text": [
"[[ 0.76484219 -0.64421769]\n",
" [ 0.64421769 0.76484219]]\n",
"[[ 1. 0.]\n",
" [ 0. 1.]]\n",
"[[ 1. 0.]\n",
" [ 0. 1.]]\n"
]
}
],
"prompt_number": 10
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# \"rotated\" quadratic form\n",
"M = array([[3.0,0.0],[0.0,1.0]])\n",
"Mr = dot(rotation(0.7).T,dot(M,rotation(0.7)))\n",
"z = apply_along_axis(lambda x:dot(x,dot(Mr,x)),0,array([x,y]))\n",
"contour(x,y,z,levels=[0.5,1,1.5])"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 11,
"text": [
"<matplotlib.contour.QuadContourSet instance at 0x3a8dcf8>"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAAT8AAAExCAYAAAAKrKXbAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\nK1fS0ttbdZ86dERwiaMcpibW
}
],
"prompt_number": 11
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"Mr"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 12,
"text": [
"array([[ 2.16996714, -0.98544973],\n",
" [-0.98544973, 1.83003286]])"
]
}
],
"prompt_number": 12
},
{
"cell_type": "heading",
"level": 1,
"metadata": {},
"source": [
"Positive Definiteness"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"Assume $A$ is a rectangular matrix with linearly independent columns.\n",
"\n",
"Claim: $A^T A$ is positive definite.\n",
"\n",
"I.e. $x^T A^T A x > 0$ for all $x\\neq 0$.\n",
"\n",
"$$x^T A^T A x = (A x)^T (A x) = ||A x||^2$$\n",
"\n",
"Clearly $\\geq 0$. $A x = 0$ only for a linear combination of columns that is zero, but that's only for $x=0$."
]
},
{
"cell_type": "heading",
"level": 1,
"metadata": {},
"source": [
"Eigenvalues and Eigenvectors"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(definition)\n",
"\n",
"$$ M x = \\lambda x $$\n",
"\n",
"If $M$ is diagonal, this is really simple: pick an axis aligned unit vector as $x$."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(eigenvalues)\n",
"\n",
"$$ M x = \\lambda x $$\n",
"\n",
"$$ (M - \\lambda I) x = 0 $$\n",
"\n",
"shifting $M$ by $\\lambda I$ makes it singular:\n",
"\n",
"$$ \\hbox{det}(M - \\lambda I) = 0$$"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(determinant)\n",
"\n",
"A determinant is an alternating multilinear function of the columns:\n",
"\n",
"$$\\hbox{det}[a_1+v,a_2,...,a_n] = \\hbox{det}[a_1,a_2,...,a_n] + \\hbox{det}[v,a_2,...,a_n]$$\n",
"\n",
"$$\\hbox{det}[a_1,a_2,...,a_n] = -\\hbox{det}[a_2,a_1,...,a_n] $$\n",
"\n",
"$$\\hbox{det}(I) = 1$$\n",
"\n",
"That actually uniquely determines it."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(determinant)\n",
"\n",
"$\\hbox{det}(M) = 0$ is equivalent to $M$ being singular.\n",
"\n",
"E.g. two identical columns:\n",
"\n",
"$$ \\hbox{det}[a,a,a_3...,a_n] = -\\hbox{det}[a,a,a_3...,a_n] $$\n",
"\n",
"Note:\n",
"\n",
"$$ \\det(AB) = \\det(A) ~ \\det(B) $$"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(2 by 2 matrices)\n",
"\n",
"$$\\hbox{det}\\left(\\begin{array}{cc}a & b \\\\\\ c & d \\\\\\ \\end{array}\\right) = ad - bc$$\n",
"\n",
"Calculate eigenvectors and eigenvalues from this:\n",
"\n",
"$$(a-\\lambda)(d-\\lambda)-bc = 0$$\n",
"\n",
"Note that some of these solutions may be complex numbers.\n",
"\n",
"If the matrix is symmetric, then the eigenvalues are always real."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(orthogonality)\n",
"\n",
"Let $M$ be symmetric and $u$ and $v$ be eigenvectors corresponding to different eigenvalues $\\lambda\\neq\\mu$. \n",
"Then $u$ and $v$ are orthogonal.\n",
"\n",
"$$ M u = \\lambda u, M v = \\mu v $$\n",
"\n",
"$$ v M^T = \\mu v^t$$\n",
"\n",
"multiply from right by u:\n",
"$$ v M^T u = \\mu v^t u$$\n",
"\n",
"apply eigenvalue equation:\n",
"$$ v \\lambda u^T = \\mu v^T u$$\n",
"\n",
"reorder:\n",
"$$ (\\lambda - \\mu) (u \\cdot v) = 0 $$\n",
"\n",
"since $\\lambda\\neq\\mu$, we must have $u\\cdot v = 0$\n"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(image coordinate mapping)\n",
"\n",
"The horizontal vector is an eigenvector of the transformation."
]
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"import urllib2\n",
"imshow(imread(urllib2.urlopen(\"http://upload.wikimedia.org/wikipedia/commons/thumb/3/3c/Mona_Lisa_eigenvector_grid.png/800px-Mona_Lisa_eigenvector_grid.png\")))"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "pyout",
"prompt_number": 13,
"text": [
"<matplotlib.image.AxesImage at 0x3d5e890>"
]
},
{
"output_type": "display_data",
"png": "iVBORw0KGgoAAAANSUhEUgAAATsAAADgCAYAAAB8UsGfAAAABHNCSVQICAgIfAhkiAAAAAlwSFlz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\niAUhnH1s7j5xswP4yle+wle+
}
],
"prompt_number": 13
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(diagonalization)\n",
"\n",
"Let $M$ be an $n\\times n$ matrix and have $n$ linearly independent eigenvectors $u_1,...,u_n$ with eigenvalues $\\lambda_1,...,\\lambda_n$\n",
"\n",
"If the eigenvectors are the columns of $S$, then\n",
"\n",
"$$ S^{-1} M S = \\Lambda = \\hbox{diag}(\\lambda_1,...,\\lambda_n) $$\n",
"\n",
"Reason:\n",
"\n",
"$$ M S = M \\left( u_1 ... u_n \\right) = \\left( M u_1 ... M u_n \\right) = \\left( \\lambda_1 u_1 ... \\lambda_n u_n \\right ) = S \\Lambda $$\n",
"\n",
"If additionally $M$ is symmetric, we can choose $S$ to be an orthogonal matrix."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(orthogonal matrix)\n",
"\n",
"An *orthogonal matrix* $Q$ is a real, square matrix where the columns and rows are orthogonal unit vectors.\n",
"\n",
"Equivalently, $Q^T = Q^{-1}$ or $QQ^T = Q^TQ = I$.\n",
"\n",
"If $Q$ is orthogonal, then $\\hbox{det}(Q) = \\pm 1$.\n",
"\n",
"If $\\hbox{det}(Q) = 1$, then $Q$ is a *rotation matrix* (otherwise it's a rotation plus reflection)."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(trace and determinant)\n",
"\n",
"The *determinant* of a matrix equals the product of the eigenvalues.\n",
"\n",
"The *trace* of a matrix is the sum of its diagonal elements and the sum of its eigenvalues."
]
},
{
"cell_type": "code",
"collapsed": false,
"input": [
"# trace and determinant\n",
"print M\n",
"print eig(Mr)\n",
"print \"trace\",trace(M)\n",
"print \"trace\",trace(Mr)\n",
"print \"det\",det(M)\n",
"print \"det\",det(Mr)"
],
"language": "python",
"metadata": {},
"outputs": [
{
"output_type": "stream",
"stream": "stdout",
"text": [
"[[ 3. 0.]\n",
" [ 0. 1.]]\n",
"(array([ 3., 1.]), array([[ 0.76484219, 0.64421769],\n",
" [-0.64421769, 0.76484219]]))\n",
"trace 4.0\n",
"trace 4.0\n",
"det 3.0\n",
"det 3.0\n"
]
}
],
"prompt_number": 17
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(proof sketches)\n",
"\n",
"determinant:\n",
"\n",
"$$ \\det(M) = \\det(S^{-1})\\det(\\Lambda)\\det(S) = 1\\cdot\\prod\\lambda_I \\cdot 1$$\n",
"\n",
"trace:\n",
"\n",
"$$ \\hbox{trace}(AB) = \\hbox{trace}(BA) $$\n",
"\n",
"$$ \\hbox{trace}(S^{-1} ~ \\Lambda S) = \\hbox{trace}(\\Lambda S ~ S^{-1}) = \\hbox{trace}(\\Lambda) $$"
]
},
{
"cell_type": "heading",
"level": 1,
"metadata": {},
"source": [
"Norms and Metrics"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(norms)\n",
"\n",
"Norms assign lengths to vectors and must satisfy:\n",
"\n",
"$$||av|| = |a| ~||v||$$\n",
"\n",
"$$||a+b|| \\leq ||a||+||b||$$\n",
"\n",
"$$||v|| = 0 \\Leftrightarrow v=0$$\n",
"\n",
"Norms imply metrics by:\n",
"\n",
"$$ d(u,v) = || u - v || $$"
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(quadratic form as norm)\n",
"\n",
"A *quadratic form* defines a norm on a vector space.\n",
"\n",
"$$ ||x||_M = \\sqrt{x^T M x} $$\n",
"\n",
"It is equivalent to a rotation and rescaling of the vector space prior to measuring distances.\n",
"\n",
"If $M = S^{-1}\\Lambda S$, then $S$ is the rotation and $\\lambda_i$ the rescaling of axis $i$."
]
},
{
"cell_type": "markdown",
"metadata": {},
"source": [
"(importance for pattern recognition)\n",
"\n",
"When you take measurements in the real world:\n",
"\n",
"- the measurements are often linear combinations of some underlying causes you are interested in\n",
"- the measurement scale of each measurement/cause is arbitrary\n",
"\n",
"Quadratic norms allow you to compensate for these effects."
]
},
{
"cell_type": "code",
"collapsed": false,
"input": [],
"language": "python",
"metadata": {},
"outputs": [],
"prompt_number": 15
}
],
"metadata": {}
}
]
}