[275] | 1 | <!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.01 Transitional//EN"> |
---|
| 2 | <html><head><meta http-equiv="Content-Type" content="text/html;charset=UTF-8"> |
---|
| 3 | <title>mixpp: Theory of ARX model estimation</title> |
---|
[290] | 4 | <link href="tabs.css" rel="stylesheet" type="text/css"> |
---|
[275] | 5 | <link href="doxygen.css" rel="stylesheet" type="text/css"> |
---|
| 6 | </head><body> |
---|
[290] | 7 | <!-- Generated by Doxygen 1.5.8 --> |
---|
[275] | 8 | <script type="text/javascript"> |
---|
| 9 | <!-- |
---|
| 10 | function changeDisplayState (e){ |
---|
| 11 | var num=this.id.replace(/[^[0-9]/g,''); |
---|
| 12 | var button=this.firstChild; |
---|
| 13 | var sectionDiv=document.getElementById('dynsection'+num); |
---|
| 14 | if (sectionDiv.style.display=='none'||sectionDiv.style.display==''){ |
---|
| 15 | sectionDiv.style.display='block'; |
---|
| 16 | button.src='open.gif'; |
---|
| 17 | }else{ |
---|
| 18 | sectionDiv.style.display='none'; |
---|
| 19 | button.src='closed.gif'; |
---|
| 20 | } |
---|
| 21 | } |
---|
| 22 | function initDynSections(){ |
---|
| 23 | var divs=document.getElementsByTagName('div'); |
---|
| 24 | var sectionCounter=1; |
---|
| 25 | for(var i=0;i<divs.length-1;i++){ |
---|
| 26 | if(divs[i].className=='dynheader'&&divs[i+1].className=='dynsection'){ |
---|
| 27 | var header=divs[i]; |
---|
| 28 | var section=divs[i+1]; |
---|
| 29 | var button=header.firstChild; |
---|
| 30 | if (button!='IMG'){ |
---|
| 31 | divs[i].insertBefore(document.createTextNode(' '),divs[i].firstChild); |
---|
| 32 | button=document.createElement('img'); |
---|
| 33 | divs[i].insertBefore(button,divs[i].firstChild); |
---|
| 34 | } |
---|
| 35 | header.style.cursor='pointer'; |
---|
| 36 | header.onclick=changeDisplayState; |
---|
| 37 | header.id='dynheader'+sectionCounter; |
---|
| 38 | button.src='closed.gif'; |
---|
| 39 | section.id='dynsection'+sectionCounter; |
---|
| 40 | section.style.display='none'; |
---|
| 41 | section.style.marginLeft='14px'; |
---|
| 42 | sectionCounter++; |
---|
| 43 | } |
---|
| 44 | } |
---|
| 45 | } |
---|
| 46 | window.onload = initDynSections; |
---|
| 47 | --> |
---|
| 48 | </script> |
---|
| 49 | <div class="navigation" id="top"> |
---|
| 50 | <div class="tabs"> |
---|
| 51 | <ul> |
---|
| 52 | <li><a href="main.html"><span>Main Page</span></a></li> |
---|
[290] | 53 | <li class="current"><a href="pages.html"><span>Related Pages</span></a></li> |
---|
[275] | 54 | <li><a href="modules.html"><span>Modules</span></a></li> |
---|
[290] | 55 | <li><a href="annotated.html"><span>Classes</span></a></li> |
---|
[275] | 56 | <li><a href="files.html"><span>Files</span></a></li> |
---|
| 57 | </ul> |
---|
| 58 | </div> |
---|
| 59 | <div class="navpath"><a class="el" href="tutorial.html">Tutorial in Bayesian estimation</a> |
---|
| 60 | </div> |
---|
| 61 | </div> |
---|
| 62 | <div class="contents"> |
---|
| 63 | <h1><a class="anchor" name="tut_arx">Theory of ARX model estimation </a></h1><p> |
---|
| 64 | The <code>ARX</code> (AutoregRessive with eXogeneous input) model is defined as follows: <p class="formulaDsp"> |
---|
[323] | 65 | <img class="formulaDsp" alt="\[ y_t = \theta' \psi_t + \rho e_t \]" src="form_107.png"> |
---|
[275] | 66 | <p> |
---|
[323] | 67 | where <img class="formulaInl" alt="$y_t$" src="form_42.png"> is the system output, <img class="formulaInl" alt="$[\theta,\rho]$" src="form_108.png"> is vector of unknown parameters, <img class="formulaInl" alt="$\psi_t$" src="form_109.png"> is an vector of data-dependent regressors, and noise <img class="formulaInl" alt="$e_t$" src="form_6.png"> is assumed to be Normal distributed <img class="formulaInl" alt="$\mathcal{N}(0,1)$" src="form_110.png">.<p> |
---|
[275] | 68 | Special cases include: <ul> |
---|
| 69 | <li>estimation of unknown mean and variance of a Gaussian density from independent samples.</li> |
---|
| 70 | </ul> |
---|
| 71 | <h2><a class="anchor" name="off"> |
---|
| 72 | Off-line estimation:</a></h2> |
---|
| 73 | This particular model belongs to the exponential family, hence it has conjugate distribution (i.e. both prior and posterior) of the Gauss-inverse-Wishart form. See [ref]<p> |
---|
| 74 | Estimation of this family can be achieved by accumulation of sufficient statistics. The sufficient statistics Gauss-inverse-Wishart density is composed of: <dl> |
---|
| 75 | <dt>Information matrix </dt> |
---|
| 76 | <dd>which is a sum of outer products <p class="formulaDsp"> |
---|
[323] | 77 | <img class="formulaDsp" alt="\[ V_t = \sum_{i=0}^{n} \left[\begin{array}{c}y_{t}\\ \psi_{t}\end{array}\right] \begin{array}{c} [y_{t}',\,\psi_{t}']\\ \\\end{array} \]" src="form_111.png"> |
---|
[275] | 78 | <p> |
---|
| 79 | </dd> |
---|
| 80 | <dt>"Degree of freedom" </dt> |
---|
| 81 | <dd>which is an accumulator of number of data records <p class="formulaDsp"> |
---|
[323] | 82 | <img class="formulaDsp" alt="\[ \nu_t = \sum_{i=0}^{n} 1 \]" src="form_112.png"> |
---|
[275] | 83 | <p> |
---|
| 84 | </dd> |
---|
| 85 | </dl> |
---|
| 86 | <h2><a class="anchor" name="on"> |
---|
| 87 | On-line estimation</a></h2> |
---|
| 88 | For online estimation with stationary parameters can be easily achieved by collecting the sufficient statistics described above recursively.<p> |
---|
[323] | 89 | Extension to non-stationaly parameters, <img class="formulaInl" alt="$ \theta_t , r_t $" src="form_113.png"> can be achieved by operation called forgetting. This is an approximation of Bayesian filtering see [Kulhavy]. The resulting algorithm is defined by manipulation of sufficient statistics: <dl> |
---|
[275] | 90 | <dt>Information matrix </dt> |
---|
| 91 | <dd>which is a sum of outer products <p class="formulaDsp"> |
---|
[323] | 92 | <img class="formulaDsp" alt="\[ V_t = \phi V_{t-1} + \left[\begin{array}{c}y_{t}\\ \psi_{t}\end{array}\right] \begin{array}{c} [y_{t}',\,\psi_{t}']\\ \\\end{array} +(1-\phi) V_0 \]" src="form_114.png"> |
---|
[275] | 93 | <p> |
---|
| 94 | </dd> |
---|
| 95 | <dt>"Degree of freedom" </dt> |
---|
| 96 | <dd>which is an accumulator of number of data records <p class="formulaDsp"> |
---|
[323] | 97 | <img class="formulaDsp" alt="\[ \nu_t = \phi \nu_{t-1} + 1 + (1-\phi) \nu_0 \]" src="form_115.png"> |
---|
[275] | 98 | <p> |
---|
| 99 | </dd> |
---|
| 100 | </dl> |
---|
[323] | 101 | where <img class="formulaInl" alt="$ \phi $" src="form_116.png"> is the forgetting factor, typically <img class="formulaInl" alt="$ \phi \in [0,1]$" src="form_117.png"> roughly corresponding to the effective length of the exponential window by relation:<p class="formulaDsp"> |
---|
| 102 | <img class="formulaDsp" alt="\[ \mathrm{win_length} = \frac{1}{1-\phi}\]" src="form_118.png"> |
---|
[275] | 103 | <p> |
---|
[323] | 104 | Hence, <img class="formulaInl" alt="$ \phi=0.9 $" src="form_119.png"> corresponds to estimation on exponential window of effective length 10 samples.<p> |
---|
| 105 | Statistics <img class="formulaInl" alt="$ V_0 , \nu_0 $" src="form_120.png"> are called alternative statistics, their role is to stabilize estimation. It is easy to show that for zero data, the statistics <img class="formulaInl" alt="$ V_t , \nu_t $" src="form_121.png"> converge to the alternative statistics.<h2><a class="anchor" name="str"> |
---|
[275] | 106 | Structure estimation</a></h2> |
---|
[323] | 107 | For this model, structure estimation is a form of model selection procedure. Specifically, we compare hypotheses that the data were generated by the full model with hypotheses that some regressors in vector <img class="formulaInl" alt="$\psi$" src="form_15.png"> are redundant. The number of possible hypotheses is then the number of all possible combinations of all regressors.<p> |
---|
| 108 | However, due to property known as nesting in exponential family, these hypotheses can be tested using only the posterior statistics. (This property does no hold for forgetting <img class="formulaInl" alt="$ \phi<1 $" src="form_122.png">). Hence, for low dimensional problems, this can be done by a tree search (method <a class="el" href="classbdm_1_1ARX.html#16b02ae03316751664c22d59d90c1e34" title="Brute force structure estimation.">bdm::ARX::structure_est()</a>). Or more sophisticated algorithm [ref Ludvik]<h2><a class="anchor" name="soft"> |
---|
[275] | 109 | Software Image</a></h2> |
---|
| 110 | Estimation of the ARX model is implemented in class <a class="el" href="classbdm_1_1ARX.html" title="Linear Autoregressive model with Gaussian noise.">bdm::ARX</a>. <ul> |
---|
| 111 | <li>models from exponential family share some properties, these are encoded in class <a class="el" href="classbdm_1_1BMEF.html" title="Estimator for Exponential family.">bdm::BMEF</a> which is the parent of ARX </li> |
---|
| 112 | <li>one of the parameters of <a class="el" href="classbdm_1_1BMEF.html" title="Estimator for Exponential family.">bdm::BMEF</a> is the forgetting factor which is stored in attribute <code>frg</code>, </li> |
---|
| 113 | <li>posterior density is stored inside the estimator in the form of <a class="el" href="classbdm_1_1egiw.html" title="Gauss-inverse-Wishart density stored in LD form.">bdm::egiw</a> </li> |
---|
| 114 | <li>references to statistics of the internal <code>egiw</code> class, i.e. attributes <code>V</code> and <code>nu</code> are established for convenience.</li> |
---|
| 115 | </ul> |
---|
| 116 | <h2><a class="anchor" name="try"> |
---|
| 117 | How to try</a></h2> |
---|
| 118 | The best way to experiment with this object is to run matlab script <code>arx_test.m</code> located in directory <code></code>./library/tutorial. See <a class="el" href="arx_ui.html">Running experiment <code>estimator</code> with ARX data fields</a> for detailed description.<p> |
---|
| 119 | <ul> |
---|
| 120 | <li>In default setup, the parameters converge to the true values as expected. </li> |
---|
| 121 | <li>Try changing the forgetting factor, field <code>estimator.frg</code>, to values <1. You should see increased lower and upper bounds on the estimates. </li> |
---|
| 122 | <li>Try different set of parameters, filed <code>system.theta</code>, you should note that poles close to zero are harder to identify. </li> |
---|
| 123 | </ul> |
---|
| 124 | </div> |
---|
[323] | 125 | <hr size="1"><address style="text-align: right;"><small>Generated on Thu Apr 23 21:06:43 2009 for mixpp by |
---|
[275] | 126 | <a href="http://www.doxygen.org/index.html"> |
---|
[290] | 127 | <img src="doxygen.png" alt="doxygen" align="middle" border="0"></a> 1.5.8 </small></address> |
---|
[275] | 128 | </body> |
---|
| 129 | </html> |
---|