docs/v1.2/group__grp__sparse__linear__solver.html - madlib-site - Git at Google

 <!-- HTML header for doxygen 1.8.4-->
 <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
 <html xmlns="http://www.w3.org/1999/xhtml">
 <head>
 <meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
 <meta http-equiv="X-UA-Compatible" content="IE=9"/>
 <meta name="generator" content="Doxygen 1.8.4"/>
 <meta name="keywords" content="madlib,postgres,greenplum,machine learning,data mining,deep learning,ensemble methods,data science,market basket analysis,affinity analysis,pca,lda,regression,elastic net,huber white,proportional hazards,k-means,latent dirichlet allocation,bayes,support vector machines,svm"/>
 <title>MADlib: Sparse Linear Systems</title>
 <link href="tabs.css" rel="stylesheet" type="text/css"/>
 <script type="text/javascript" src="jquery.js"></script>
 <script type="text/javascript" src="dynsections.js"></script>
 <link href="navtree.css" rel="stylesheet" type="text/css"/>
 <script type="text/javascript" src="resize.js"></script>
 <script type="text/javascript" src="navtree.js"></script>
 <script type="text/javascript">
   $(document).ready(initResizable);
   $(window).load(resizeHeight);
 </script>
 <link href="search/search.css" rel="stylesheet" type="text/css"/>
 <script type="text/javascript" src="search/search.js"></script>
 <script type="text/javascript">
   $(document).ready(function() { searchBox.OnSelectItem(0); });
 </script>
 <script type="text/x-mathjax-config">
   MathJax.Hub.Config({
     extensions: ["tex2jax.js", "TeX/AMSmath.js", "TeX/AMSsymbols.js"],
     jax: ["input/TeX","output/HTML-CSS"],
 });
 </script><script src="../mathjax/MathJax.js"></script>
 <link href="doxygen.css" rel="stylesheet" type="text/css" />
 <link href="madlib_extra.css" rel="stylesheet" type="text/css"/>
 <!-- google analytics -->
 <script>
   (function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]||function(){
   (i[r].q=i[r].q||[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),
   m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)
   })(window,document,'script','//www.google-analytics.com/analytics.js','ga');
   ga('create', 'UA-45382226-1', 'auto');
   ga('send', 'pageview');
 </script>
 </head>
 <body>
 <div id="top"><!-- do not remove this div, it is closed by doxygen! -->
 <div id="titlearea">
 <table cellspacing="0" cellpadding="0">
  <tbody>
  <tr style="height: 56px;">
   <td style="padding-left: 0.5em;">
    <div id="projectname">MADlib
    &#160;<span id="projectnumber">1.2</span> <span style="font-size:10pt; font-style:italic"><a href="../latest/./group__grp__sparse__linear__solver.html"> A newer version is available</a></span>
    </div>
    <div id="projectbrief">User Documentation</div>
   </td>
   <!--BEGIN VERSIONS LINKS-->
     <td style="padding-left: 0.5em;">
       <div class="versionlist"><ul>
         <li class="head">More versions:</li>
         <li><a href="../v1.1/index.html">v1.1</li>
         <li><a href="../v1.0/index.html">v1.0</li>
         <li><a href="../v0.7/index.html">v0.7</li>
         <li><a href="../v0.5/index.html">v0.5</li></ul>
       </div>
     </td>
    <td>        <div id="MSearchBox" class="MSearchBoxInactive">
         <span class="left">
           <img id="MSearchSelect" src="search/mag_sel.png"
                onmouseover="return searchBox.OnSearchSelectShow()"
                onmouseout="return searchBox.OnSearchSelectHide()"
                alt=""/>
           <input type="text" id="MSearchField" value="Search" accesskey="S"
                onfocus="searchBox.OnSearchFieldFocus(true)"
                onblur="searchBox.OnSearchFieldFocus(false)"
                onkeyup="searchBox.OnSearchFieldChange(event)"/>
           </span><span class="right">
             <a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
           </span>
         </div>
 </td>
  </tr>
  </tbody>
 </table>
 </div>
 <!-- end header part -->
 <!-- Generated by Doxygen 1.8.4 -->
 <script type="text/javascript">
 var searchBox = new SearchBox("searchBox", "search",false,'Search');
 </script>
 </div><!-- top -->
 <div id="side-nav" class="ui-resizable side-nav-resizable">
   <div id="nav-tree">
     <div id="nav-tree-contents">
       <div id="nav-sync" class="sync"></div>
     </div>
   </div>
   <div id="splitbar" style="-moz-user-select:none;"
        class="ui-resizable-handle">
   </div>
 </div>
 <script type="text/javascript">
 $(document).ready(function(){initNavTree('group__grp__sparse__linear__solver.html','');});
 </script>
 <div id="doc-content">
 <!-- window showing the filter options -->
 <div id="MSearchSelectWindow"
      onmouseover="return searchBox.OnSearchSelectShow()"
      onmouseout="return searchBox.OnSearchSelectHide()"
      onkeydown="return searchBox.OnSearchSelectKey(event)">
 <a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(0)"><span class="SelectionMark">&#160;</span>All</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(1)"><span class="SelectionMark">&#160;</span>Files</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(2)"><span class="SelectionMark">&#160;</span>Functions</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(3)"><span class="SelectionMark">&#160;</span>Variables</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(4)"><span class="SelectionMark">&#160;</span>Groups</a></div>

 <!-- iframe showing the search results (closed by default) -->
 <div id="MSearchResultsWindow">
 <iframe src="javascript:void(0)" frameborder="0"
         name="MSearchResults" id="MSearchResults">
 </iframe>
 </div>

 <div class="header">
   <div class="headertitle">
 <div class="title">Sparse Linear Systems<div class="ingroups"><a class="el" href="group__grp__linear__solver.html">Linear Systems</a></div></div>  </div>
 </div><!--header-->
 <div class="contents">
 <div class="toc"><b>Contents</b> </p>
 <ul>
 <li class="level1">
 <a href="#sls_about">About</a> </li>
 <li class="level1">
 <a href="#sls_online_help">Online Help</a> </li>
 <li class="level1">
 <a href="#sls_usage">Function Syntax</a> </li>
 <li class="level1">
 <a href="#sls_args">Arguments</a> </li>
 <li class="level1">
 <a href="#sls_opt_params">Optimizer Parameters</a> </li>
 <li class="level1">
 <a href="#sls_output">Output Tables</a> </li>
 <li class="level1">
 <a href="#sls_examples">Examples</a> </li>
 </ul>
 </div><p><a class="anchor" id="sls_about"></a></p>
 <dl class="section user"><dt>About</dt><dd></dd></dl>
 <p>The sparse linear systems module implements solution methods for systems of a consistent linear equations. Systems of linear equations take the form: </p>
 <p class="formulaDsp">
 \[ Ax = b \]
 </p>
 <p>where \(x \in \mathbb{R}^{n}\), \(A \in \mathbb{R}^{m \times n} \) and \(b \in \mathbb{R}^{m}\). This module accepts sparse matrix input formats for \(A\) and \(b\). We assume that there are no rows of \(A\) where all elements are zero.</p>
 <dl class="section note"><dt>Note</dt><dd>Algorithms with fail if there is an row of the input matrix containing all zeros.</dd></dl>
 <p>The algorithms implemented in this module can handle large sparse square linear systems. Currently, the algorithms implemented in this module solve the linear system using direct or iterative methods.</p>
 <p><a class="anchor" id="sls_online_help"></a></p>
 <dl class="section user"><dt>Online Help</dt><dd></dd></dl>
 <p>View short help messages using the following statements: </p>
 <pre class="fragment">-- Summary of sparse linear systems
 SELECT madlib.linear_solver_sparse();

 -- Function syntax and output table format
 SELECT madlib.linear_solver_sparse('usage');

 -- Syntax for direct methods
 SELECT madlib.linear_solver_sparse('direct');

 -- Syntax for iterative methods
 SELECT madlib.linear_solver_sparse('iterative');
 </pre><p><a class="anchor" id="sls_usage"></a></p>
 <dl class="section user"><dt>Usage</dt><dd></dd></dl>
 <pre class="fragment">linear_solver_sparse(tbl_source_lhs, tbl_source_rhs, tbl_result,
                      lhs_row_id, lhs_col_id, lhs_value,
                      rhs_row_id, rhs_value,
                      grouping_cols := NULL,
                      optimizer := 'direct',
                      optimizer_params := 'algorithm = llt');
 </pre><p><a class="anchor" id="sls_args"></a></p>
 <dl class="section user"><dt>Arguments</dt><dd></dd></dl>
 <dl class="arglist">
 <dt>tbl_source_lhs </dt>
 <dd><p class="startdd">The name of the table containing the left hand side matrix. For the LHS matrix, the input data is expected to be of the following form: </p>
 <pre class="fragment">{TABLE|VIEW} &lt;em&gt;sourceName&lt;/em&gt; (
     ...
     &lt;em&gt;row_id&lt;/em&gt; FLOAT8,
     &lt;em&gt;col_id&lt;/em&gt; FLOAT8,
     &lt;em&gt;value&lt;/em&gt; FLOAT8,
     ...
 )</pre><p> Here, each row represents a single equation using. The <em> rhs </em> refers to the right hand side of the equations while the <em> lhs</em> refers to the multipliers on the variables on the left hand side of the same equations. </p>
 <p class="enddd"></p>
 </dd>
 <dt>tbl_source_rhs </dt>
 <dd><p class="startdd">TEXT. The name of the table containing the right hand side vector. For the RHS matrix, the input data is expected to be of the following form: </p>
 <pre class="fragment">{TABLE|VIEW} &lt;em&gt;sourceName&lt;/em&gt; (
     ...
     &lt;em&gt;row_id&lt;/em&gt; FLOAT8,
     &lt;em&gt;value&lt;/em&gt; FLOAT8
     ...
 )</pre><p> Here, each row represents a single equation using. The <em> rhs </em> refers to the right hand side of the equations while the <em> lhs</em> refers to the multipliers on the variables on the left hand side of the same equations.</p>
 <p>Output is stored in the <em>tbl_result</em> </p>
 <pre class="fragment">    tbl_result      |   Data Types
 --------------------|---------------------
 solution            | DOUBLE PRECISION[]
 residual_norm       | DOUBLE PRECISIOn
 iters               | INTEGER
 </pre> <p class="enddd"></p>
 </dd>
 <dt>tbl_result </dt>
 <dd><p class="startdd">TEXT. The name of the table where the output is saved.</p>
 <p class="enddd"></p>
 </dd>
 <dt>lhs_row_id </dt>
 <dd>TEXT. The name of the column storing the 'row id' of the equations. <dl class="section note"><dt>Note</dt><dd>For a system with N equations, the row_id's must be a continuous range of integers from \( 0 \ldots n-1 \).</dd></dl>
 </dd>
 <dt>lhs_col_id </dt>
 <dd><p class="startdd">TEXT. The name of the column (in tbl_source_lhs) storing the 'col id' of the equations.</p>
 <p class="enddd"></p>
 </dd>
 <dt>lhs_value </dt>
 <dd><p class="startdd">TEXT. The name of the column (in tbl_source_lhs) storing the 'value' of the equations.</p>
 <p class="enddd"></p>
 </dd>
 <dt>rhs_row_id </dt>
 <dd><p class="startdd">TEXT. The name of the column (in tbl_source_rhs) storing the 'col id' of the equations.</p>
 <p class="enddd"></p>
 </dd>
 <dt>rhs_value </dt>
 <dd><p class="startdd">TEXT. The name of the column (in tbl_source_rhs) storing the 'value' of the equations.</p>
 <p class="enddd"></p>
 </dd>
 <dt>num_vars </dt>
 <dd><p class="startdd">INTEGER. The number of variables in the linear system. equations.</p>
 <p class="enddd"></p>
 </dd>
 <dt>grouping_col (optional)  </dt>
 <dd>TEXT. Group by column names. Default: NULL. <dl class="section note"><dt>Note</dt><dd>The grouping feature is currently not implemented and this parameter is only a placeholder.</dd></dl>
 </dd>
 <dt>optimizer (optional)  </dt>
 <dd><p class="startdd">TEXT. Type of optimizer. Default: 'direct'.</p>
 <p class="enddd"></p>
 </dd>
 <dt>optimizer_params (optional)  </dt>
 <dd>TEXT. Optimizer specific parameters. Default: NULL. </dd>
 </dl>
 <p><a class="anchor" id="sls_opt_params"></a></p>
 <dl class="section user"><dt>Optimizer Parameters</dt><dd></dd></dl>
 <p>For each optimizer, there are specific parameters that can be tuned for better performance.</p>
 <dl class="arglist">
 <dt>algorithm (default: ldlt) </dt>
 <dd><p class="startdd"></p>
 <p>There are several algorithms that can be classified as 'direct' methods of solving linear systems. Madlib functions provide various algorithmic options available for users.</p>
 <p>The following table provides a guideline on the choice of algorithm based on conditions on the A matrix, speed of the algorithms and numerical stability.</p>
 <pre class="fragment">    Algorithm          | Contitions on A  | Speed | Memory
     ----------------------------------------------------------
     llt                | Sym. Pos Def     |  ++   |  ---
     ldlt               | Sym. Pos Def     |  ++   |  ---

     For speed '++' is faster than '+', which is faster than '-'.
     For accuracy '+++' is better than '++'.
     For memory, '-' uses less memory than '--'.

     Note: ldlt is often preferred over llt
 </pre><p>There are several algorithms that can be classified as 'iterative' methods of solving linear systems. Madlib functions provide various algorithmic options available for users.</p>
 <p>The following table provides a guideline on the choice of algorithm based on conditions on the A matrix, speed of the algorithms and numerical stability.</p>
 <pre class="fragment">    Algorithm            | Contitions on A  | Speed | Memory | Convergence
     ----------------------------------------------------------------------
     cg-mem               | Sym. Pos Def     |  +++  |   -    |    ++
     bicgstab-mem         | Square           |  ++   |   -    |    +
     precond-cg-mem       | Sym. Pos Def     |  ++   |   -    |    +++
     precond-bicgstab-mem | Square           |  +    |   -    |    ++

     For memory, '-' uses less memory than '--'.
     For speed, '++' is faster than '+'.
 </pre><p>Details:</p>
 <ol type="1">
 <li><b>cg-mem: </b> In memory conjugate gradient with diagonal preconditioners.</li>
 <li><b>bicgstab-mem: </b> Bi-conjugate gradient (equivalent to performing CG on the least squares formulation of Ax=b) with incomplete LU preconditioners.</li>
 <li><b>precond-cg-mem:</b> In memory conjugate gradient with diagonal preconditioners.</li>
 <li><b>bicgstab-mem: </b> Bi-conjugate gradient (equivalent to performing CG on the least squares formulation of Ax=b) with incomplete LU preconditioners. </li>
 </ol>
 <p class="enddd"></p>
 </dd>
 <dt>toler (default: 1e-5) </dt>
 <dd><p class="startdd">Termination tolerance (applicable only for iterative methods) which determines the stopping criterion (with respect to residual norm) for iterative methods. </p>
 <p class="enddd"></p>
 </dd>
 </dl>
 <p><a class="anchor" id="sls_output"></a></p>
 <dl class="section user"><dt>Output statistics</dt><dd></dd></dl>
 <dl class="arglist">
 <dt>solution </dt>
 <dd><p class="startdd">FLOAT8[]. The solution is an array with the variables in the same order as that provided as input in the 'left_hand_side' column name of the 'source_table' </p>
 <p class="enddd"></p>
 </dd>
 <dt>residual_norm </dt>
 <dd><p class="startdd">FLOAT8. Scaled residual norm, defined as \( \frac{|Ax - b|}{|b|} \). This value is an indication of the accuracy of the solution. </p>
 <p class="enddd"></p>
 </dd>
 <dt>iters </dt>
 <dd>INTEGER. Number of iterations required by the algorithm (only applicable for iterative algorithms) . The output is NULL for 'direct' methods.  </dd>
 </dl>
 <p><a class="anchor" id="sls_examples"></a></p>
 <dl class="section user"><dt>Examples</dt><dd><ol type="1">
 <li>Create the sample data set: <pre class="fragment">sql&gt;  DROP TABLE IF EXISTS sparse_linear_systems_lhs;
       CREATE TABLE sparse_linear_systems_lhs (
           rid INTEGER NOT NULL,
           cid  INTEGER,
           val DOUBLE PRECISION
       );

       DROP TABLE IF EXISTS sparse_linear_systems_rhs;
       CREATE TABLE sparse_linear_systems_rhs (
           rid INTEGER NOT NULL,
           val DOUBLE PRECISION
       );


       INSERT INTO sparse_linear_systems_lhs(rid, cid, val) VALUES
       (0, 0, 1),
       (1, 1, 1),
       (2, 2, 1),
       (3, 3, 1);

       INSERT INTO sparse_linear_systems_rhs(rid, val) VALUES
       (0, 10),
       (1, 20),
       (2, 30);
 </pre></li>
 <li>Solve the linear systems with default parameters <pre class="fragment">sql&gt; SELECT madlib.linear_solver_sparse(
                                  'sparse_linear_systems_lhs',
                                  'sparse_linear_systems_rhs',
                                  'output_table',
                                  'rid',
                                  'cid',
                                  'val',
                                  'rid',
                                  'val',
                                  4);
 </pre></li>
 <li>Obtain the output from the output table <pre class="fragment">sql&gt; SELECT * FROM output_table;
 --------------------+-------------------------------------
 solution            | {10,20,30,0}
 residual_norm       | 0
 iters               | NULL
 </pre></li>
 <li>Chose a different algorithm than the default algorithm <pre class="fragment">drop table if exists result_table;

 SELECT madlib.linear_solver_sparse(
                                  'sparse_linear_systems_lhs',
                                  'sparse_linear_systems_rhs',
                                  'output_table',
                                  'rid',
                                  'cid',
                                  'val',
                                  'rid',
                                  'val',
                                  4,
                                  NULL,
                                  'direct',
                                  'algorithm=llt');

 -# Chose a different algorithm than the default algorithm
 @verbatim
 drop table if exists result_table;
 madlib.linear_solver_sparse(
                              'sparse_linear_systems_lhs',
                              'sparse_linear_systems_rhs',
                              'output_table',
                              'rid',
                              'cid',
                              'val',
                              'rid',
                              'val',
                              4,
                              NULL,
                              'iterative',
                              'algorithm=cg-mem, toler=1e-5');
 </pre></li>
 </ol>
 </dd></dl>
 <dl class="section see"><dt>See Also</dt><dd>File sparse_linear_sytems.sql_in documenting the SQL functions. </dd></dl>
 </div><!-- contents -->
 </div><!-- doc-content -->
 <!-- start footer part -->
 <div id="nav-path" class="navpath"><!-- id is needed for treeview function! -->
   <ul>
     <li class="footer">Generated on Thu Jan 9 2014 20:35:40 for MADlib by
     <a href="http://www.doxygen.org/index.html">
     <img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.4 </li>
   </ul>
 </div>
 </body>
 </html>
	<!-- HTML header for doxygen 1.8.4-->
	<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-transitional.dtd">
	<html xmlns="http://www.w3.org/1999/xhtml">
	<head>
	<meta http-equiv="Content-Type" content="text/xhtml;charset=UTF-8"/>
	<meta http-equiv="X-UA-Compatible" content="IE=9"/>
	<meta name="generator" content="Doxygen 1.8.4"/>
	<meta name="keywords" content="madlib,postgres,greenplum,machine learning,data mining,deep learning,ensemble methods,data science,market basket analysis,affinity analysis,pca,lda,regression,elastic net,huber white,proportional hazards,k-means,latent dirichlet allocation,bayes,support vector machines,svm"/>
	<title>MADlib: Sparse Linear Systems</title>
	<link href="tabs.css" rel="stylesheet" type="text/css"/>
	<script type="text/javascript" src="jquery.js"></script>
	<script type="text/javascript" src="dynsections.js"></script>
	<link href="navtree.css" rel="stylesheet" type="text/css"/>
	<script type="text/javascript" src="resize.js"></script>
	<script type="text/javascript" src="navtree.js"></script>
	<script type="text/javascript">
	$(document).ready(initResizable);
	$(window).load(resizeHeight);
	</script>
	<link href="search/search.css" rel="stylesheet" type="text/css"/>
	<script type="text/javascript" src="search/search.js"></script>
	<script type="text/javascript">
	$(document).ready(function() { searchBox.OnSelectItem(0); });
	</script>
	<script type="text/x-mathjax-config">
	MathJax.Hub.Config({
	extensions: ["tex2jax.js", "TeX/AMSmath.js", "TeX/AMSsymbols.js"],
	jax: ["input/TeX","output/HTML-CSS"],
	});
	</script><script src="../mathjax/MathJax.js"></script>
	<link href="doxygen.css" rel="stylesheet" type="text/css" />
	<link href="madlib_extra.css" rel="stylesheet" type="text/css"/>
	<!-- google analytics -->
	<script>
	(function(i,s,o,g,r,a,m){i['GoogleAnalyticsObject']=r;i[r]=i[r]\|\|function(){
	(i[r].q=i[r].q\|\|[]).push(arguments)},i[r].l=1*new Date();a=s.createElement(o),
	m=s.getElementsByTagName(o)[0];a.async=1;a.src=g;m.parentNode.insertBefore(a,m)
	})(window,document,'script','//www.google-analytics.com/analytics.js','ga');
	ga('create', 'UA-45382226-1', 'auto');
	ga('send', 'pageview');
	</script>
	</head>
	<body>
	<div id="top"><!-- do not remove this div, it is closed by doxygen! -->
	<div id="titlearea">
	<table cellspacing="0" cellpadding="0">
	<tbody>
	<tr style="height: 56px;">
	<td style="padding-left: 0.5em;">
	<div id="projectname">MADlib
	<span id="projectnumber">1.2</span> <span style="font-size:10pt; font-style:italic"><a href="../latest/./group__grp__sparse__linear__solver.html"> A newer version is available</a></span>
	</div>
	<div id="projectbrief">User Documentation</div>
	</td>
	<!--BEGIN VERSIONS LINKS-->
	<td style="padding-left: 0.5em;">
	<div class="versionlist"><ul>
	<li class="head">More versions:</li>
	<li><a href="../v1.1/index.html">v1.1</li>
	<li><a href="../v1.0/index.html">v1.0</li>
	<li><a href="../v0.7/index.html">v0.7</li>
	<li><a href="../v0.5/index.html">v0.5</li></ul>
	</div>
	</td>
	<td> <div id="MSearchBox" class="MSearchBoxInactive">
	<span class="left">
	<img id="MSearchSelect" src="search/mag_sel.png"
	onmouseover="return searchBox.OnSearchSelectShow()"
	onmouseout="return searchBox.OnSearchSelectHide()"
	alt=""/>
	<input type="text" id="MSearchField" value="Search" accesskey="S"
	onfocus="searchBox.OnSearchFieldFocus(true)"
	onblur="searchBox.OnSearchFieldFocus(false)"
	onkeyup="searchBox.OnSearchFieldChange(event)"/>
	</span><span class="right">
	<a id="MSearchClose" href="javascript:searchBox.CloseResultsWindow()"><img id="MSearchCloseImg" border="0" src="search/close.png" alt=""/></a>
	</span>
	</div>
	</td>
	</tr>
	</tbody>
	</table>
	</div>
	<!-- end header part -->
	<!-- Generated by Doxygen 1.8.4 -->
	<script type="text/javascript">
	var searchBox = new SearchBox("searchBox", "search",false,'Search');
	</script>
	</div><!-- top -->
	<div id="side-nav" class="ui-resizable side-nav-resizable">
	<div id="nav-tree">
	<div id="nav-tree-contents">
	<div id="nav-sync" class="sync"></div>
	</div>
	</div>
	<div id="splitbar" style="-moz-user-select:none;"
	class="ui-resizable-handle">
	</div>
	</div>
	<script type="text/javascript">
	$(document).ready(function(){initNavTree('group__grp__sparse__linear__solver.html','');});
	</script>
	<div id="doc-content">
	<!-- window showing the filter options -->
	<div id="MSearchSelectWindow"
	onmouseover="return searchBox.OnSearchSelectShow()"
	onmouseout="return searchBox.OnSearchSelectHide()"
	onkeydown="return searchBox.OnSearchSelectKey(event)">
	<a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(0)"><span class="SelectionMark"> </span>All</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(1)"><span class="SelectionMark"> </span>Files</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(2)"><span class="SelectionMark"> </span>Functions</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(3)"><span class="SelectionMark"> </span>Variables</a><a class="SelectItem" href="javascript:void(0)" onclick="searchBox.OnSelectItem(4)"><span class="SelectionMark"> </span>Groups</a></div>

	<!-- iframe showing the search results (closed by default) -->
	<div id="MSearchResultsWindow">
	<iframe src="javascript:void(0)" frameborder="0"
	name="MSearchResults" id="MSearchResults">
	</iframe>
	</div>

	<div class="header">
	<div class="headertitle">
	<div class="title">Sparse Linear Systems<div class="ingroups"><a class="el" href="group__grp__linear__solver.html">Linear Systems</a></div></div> </div>
	</div><!--header-->
	<div class="contents">
	<div class="toc"><b>Contents</b> </p>
	<ul>
	<li class="level1">
	<a href="#sls_about">About</a> </li>
	<li class="level1">
	<a href="#sls_online_help">Online Help</a> </li>
	<li class="level1">
	<a href="#sls_usage">Function Syntax</a> </li>
	<li class="level1">
	<a href="#sls_args">Arguments</a> </li>
	<li class="level1">
	<a href="#sls_opt_params">Optimizer Parameters</a> </li>
	<li class="level1">
	<a href="#sls_output">Output Tables</a> </li>
	<li class="level1">
	<a href="#sls_examples">Examples</a> </li>
	</ul>
	</div><p><a class="anchor" id="sls_about"></a></p>
	<dl class="section user"><dt>About</dt><dd></dd></dl>
	<p>The sparse linear systems module implements solution methods for systems of a consistent linear equations. Systems of linear equations take the form: </p>
	<p class="formulaDsp">
	\[ Ax = b \]
	</p>
	<p>where \(x \in \mathbb{R}^{n}\), \(A \in \mathbb{R}^{m \times n} \) and \(b \in \mathbb{R}^{m}\). This module accepts sparse matrix input formats for \(A\) and \(b\). We assume that there are no rows of \(A\) where all elements are zero.</p>
	<dl class="section note"><dt>Note</dt><dd>Algorithms with fail if there is an row of the input matrix containing all zeros.</dd></dl>
	<p>The algorithms implemented in this module can handle large sparse square linear systems. Currently, the algorithms implemented in this module solve the linear system using direct or iterative methods.</p>
	<p><a class="anchor" id="sls_online_help"></a></p>
	<dl class="section user"><dt>Online Help</dt><dd></dd></dl>
	<p>View short help messages using the following statements: </p>
	<pre class="fragment">-- Summary of sparse linear systems
	SELECT madlib.linear_solver_sparse();

	-- Function syntax and output table format
	SELECT madlib.linear_solver_sparse('usage');

	-- Syntax for direct methods
	SELECT madlib.linear_solver_sparse('direct');

	-- Syntax for iterative methods
	SELECT madlib.linear_solver_sparse('iterative');
	</pre><p><a class="anchor" id="sls_usage"></a></p>
	<dl class="section user"><dt>Usage</dt><dd></dd></dl>
	<pre class="fragment">linear_solver_sparse(tbl_source_lhs, tbl_source_rhs, tbl_result,
	lhs_row_id, lhs_col_id, lhs_value,
	rhs_row_id, rhs_value,
	grouping_cols := NULL,
	optimizer := 'direct',
	optimizer_params := 'algorithm = llt');
	</pre><p><a class="anchor" id="sls_args"></a></p>
	<dl class="section user"><dt>Arguments</dt><dd></dd></dl>
	<dl class="arglist">
	<dt>tbl_source_lhs </dt>
	<dd><p class="startdd">The name of the table containing the left hand side matrix. For the LHS matrix, the input data is expected to be of the following form: </p>
	<pre class="fragment">{TABLE\|VIEW} <em>sourceName</em> (
	...
	<em>row_id</em> FLOAT8,
	<em>col_id</em> FLOAT8,
	<em>value</em> FLOAT8,
	...
	)</pre><p> Here, each row represents a single equation using. The <em> rhs </em> refers to the right hand side of the equations while the <em> lhs</em> refers to the multipliers on the variables on the left hand side of the same equations. </p>
	<p class="enddd"></p>
	</dd>
	<dt>tbl_source_rhs </dt>
	<dd><p class="startdd">TEXT. The name of the table containing the right hand side vector. For the RHS matrix, the input data is expected to be of the following form: </p>
	<pre class="fragment">{TABLE\|VIEW} <em>sourceName</em> (
	...
	<em>row_id</em> FLOAT8,
	<em>value</em> FLOAT8
	...
	)</pre><p> Here, each row represents a single equation using. The <em> rhs </em> refers to the right hand side of the equations while the <em> lhs</em> refers to the multipliers on the variables on the left hand side of the same equations.</p>
	<p>Output is stored in the <em>tbl_result</em> </p>
	<pre class="fragment"> tbl_result \| Data Types
	--------------------\|---------------------
	solution \| DOUBLE PRECISION[]
	residual_norm \| DOUBLE PRECISIOn
	iters \| INTEGER
	</pre> <p class="enddd"></p>
	</dd>
	<dt>tbl_result </dt>
	<dd><p class="startdd">TEXT. The name of the table where the output is saved.</p>
	<p class="enddd"></p>
	</dd>
	<dt>lhs_row_id </dt>
	<dd>TEXT. The name of the column storing the 'row id' of the equations. <dl class="section note"><dt>Note</dt><dd>For a system with N equations, the row_id's must be a continuous range of integers from \( 0 \ldots n-1 \).</dd></dl>
	</dd>
	<dt>lhs_col_id </dt>
	<dd><p class="startdd">TEXT. The name of the column (in tbl_source_lhs) storing the 'col id' of the equations.</p>
	<p class="enddd"></p>
	</dd>
	<dt>lhs_value </dt>
	<dd><p class="startdd">TEXT. The name of the column (in tbl_source_lhs) storing the 'value' of the equations.</p>
	<p class="enddd"></p>
	</dd>
	<dt>rhs_row_id </dt>
	<dd><p class="startdd">TEXT. The name of the column (in tbl_source_rhs) storing the 'col id' of the equations.</p>
	<p class="enddd"></p>
	</dd>
	<dt>rhs_value </dt>
	<dd><p class="startdd">TEXT. The name of the column (in tbl_source_rhs) storing the 'value' of the equations.</p>
	<p class="enddd"></p>
	</dd>
	<dt>num_vars </dt>
	<dd><p class="startdd">INTEGER. The number of variables in the linear system. equations.</p>
	<p class="enddd"></p>
	</dd>
	<dt>grouping_col (optional) </dt>
	<dd>TEXT. Group by column names. Default: NULL. <dl class="section note"><dt>Note</dt><dd>The grouping feature is currently not implemented and this parameter is only a placeholder.</dd></dl>
	</dd>
	<dt>optimizer (optional) </dt>
	<dd><p class="startdd">TEXT. Type of optimizer. Default: 'direct'.</p>
	<p class="enddd"></p>
	</dd>
	<dt>optimizer_params (optional) </dt>
	<dd>TEXT. Optimizer specific parameters. Default: NULL. </dd>
	</dl>
	<p><a class="anchor" id="sls_opt_params"></a></p>
	<dl class="section user"><dt>Optimizer Parameters</dt><dd></dd></dl>
	<p>For each optimizer, there are specific parameters that can be tuned for better performance.</p>
	<dl class="arglist">
	<dt>algorithm (default: ldlt) </dt>
	<dd><p class="startdd"></p>
	<p>There are several algorithms that can be classified as 'direct' methods of solving linear systems. Madlib functions provide various algorithmic options available for users.</p>
	<p>The following table provides a guideline on the choice of algorithm based on conditions on the A matrix, speed of the algorithms and numerical stability.</p>
	<pre class="fragment"> Algorithm \| Contitions on A \| Speed \| Memory
	----------------------------------------------------------
	llt \| Sym. Pos Def \| ++ \| ---
	ldlt \| Sym. Pos Def \| ++ \| ---

	For speed '++' is faster than '+', which is faster than '-'.
	For accuracy '+++' is better than '++'.
	For memory, '-' uses less memory than '--'.

	Note: ldlt is often preferred over llt
	</pre><p>There are several algorithms that can be classified as 'iterative' methods of solving linear systems. Madlib functions provide various algorithmic options available for users.</p>
	<p>The following table provides a guideline on the choice of algorithm based on conditions on the A matrix, speed of the algorithms and numerical stability.</p>
	<pre class="fragment"> Algorithm \| Contitions on A \| Speed \| Memory \| Convergence
	----------------------------------------------------------------------
	cg-mem \| Sym. Pos Def \| +++ \| - \| ++
	bicgstab-mem \| Square \| ++ \| - \| +
	precond-cg-mem \| Sym. Pos Def \| ++ \| - \| +++
	precond-bicgstab-mem \| Square \| + \| - \| ++

	For memory, '-' uses less memory than '--'.
	For speed, '++' is faster than '+'.
	</pre><p>Details:</p>
	<ol type="1">
	<li><b>cg-mem: </b> In memory conjugate gradient with diagonal preconditioners.</li>
	<li><b>bicgstab-mem: </b> Bi-conjugate gradient (equivalent to performing CG on the least squares formulation of Ax=b) with incomplete LU preconditioners.</li>
	<li><b>precond-cg-mem:</b> In memory conjugate gradient with diagonal preconditioners.</li>
	<li><b>bicgstab-mem: </b> Bi-conjugate gradient (equivalent to performing CG on the least squares formulation of Ax=b) with incomplete LU preconditioners. </li>
	</ol>
	<p class="enddd"></p>
	</dd>
	<dt>toler (default: 1e-5) </dt>
	<dd><p class="startdd">Termination tolerance (applicable only for iterative methods) which determines the stopping criterion (with respect to residual norm) for iterative methods. </p>
	<p class="enddd"></p>
	</dd>
	</dl>
	<p><a class="anchor" id="sls_output"></a></p>
	<dl class="section user"><dt>Output statistics</dt><dd></dd></dl>
	<dl class="arglist">
	<dt>solution </dt>
	<dd><p class="startdd">FLOAT8[]. The solution is an array with the variables in the same order as that provided as input in the 'left_hand_side' column name of the 'source_table' </p>
	<p class="enddd"></p>
	</dd>
	<dt>residual_norm </dt>
	<dd><p class="startdd">FLOAT8. Scaled residual norm, defined as \( \frac{\|Ax - b\|}{\|b\|} \). This value is an indication of the accuracy of the solution. </p>
	<p class="enddd"></p>
	</dd>
	<dt>iters </dt>
	<dd>INTEGER. Number of iterations required by the algorithm (only applicable for iterative algorithms) . The output is NULL for 'direct' methods. </dd>
	</dl>
	<p><a class="anchor" id="sls_examples"></a></p>
	<dl class="section user"><dt>Examples</dt><dd><ol type="1">
	<li>Create the sample data set: <pre class="fragment">sql> DROP TABLE IF EXISTS sparse_linear_systems_lhs;
	CREATE TABLE sparse_linear_systems_lhs (
	rid INTEGER NOT NULL,
	cid INTEGER,
	val DOUBLE PRECISION
	);

	DROP TABLE IF EXISTS sparse_linear_systems_rhs;
	CREATE TABLE sparse_linear_systems_rhs (
	rid INTEGER NOT NULL,
	val DOUBLE PRECISION
	);


	INSERT INTO sparse_linear_systems_lhs(rid, cid, val) VALUES
	(0, 0, 1),
	(1, 1, 1),
	(2, 2, 1),
	(3, 3, 1);

	INSERT INTO sparse_linear_systems_rhs(rid, val) VALUES
	(0, 10),
	(1, 20),
	(2, 30);
	</pre></li>
	<li>Solve the linear systems with default parameters <pre class="fragment">sql> SELECT madlib.linear_solver_sparse(
	'sparse_linear_systems_lhs',
	'sparse_linear_systems_rhs',
	'output_table',
	'rid',
	'cid',
	'val',
	'rid',
	'val',
	4);
	</pre></li>
	<li>Obtain the output from the output table <pre class="fragment">sql> SELECT * FROM output_table;
	--------------------+-------------------------------------
	solution \| {10,20,30,0}
	residual_norm \| 0
	iters \| NULL
	</pre></li>
	<li>Chose a different algorithm than the default algorithm <pre class="fragment">drop table if exists result_table;

	SELECT madlib.linear_solver_sparse(
	'sparse_linear_systems_lhs',
	'sparse_linear_systems_rhs',
	'output_table',
	'rid',
	'cid',
	'val',
	'rid',
	'val',
	4,
	NULL,
	'direct',
	'algorithm=llt');

	-# Chose a different algorithm than the default algorithm
	@verbatim
	drop table if exists result_table;
	madlib.linear_solver_sparse(
	'sparse_linear_systems_lhs',
	'sparse_linear_systems_rhs',
	'output_table',
	'rid',
	'cid',
	'val',
	'rid',
	'val',
	4,
	NULL,
	'iterative',
	'algorithm=cg-mem, toler=1e-5');
	</pre></li>
	</ol>
	</dd></dl>
	<dl class="section see"><dt>See Also</dt><dd>File sparse_linear_sytems.sql_in documenting the SQL functions. </dd></dl>
	</div><!-- contents -->
	</div><!-- doc-content -->
	<!-- start footer part -->
	<div id="nav-path" class="navpath"><!-- id is needed for treeview function! -->
	<ul>
	<li class="footer">Generated on Thu Jan 9 2014 20:35:40 for MADlib by
	<a href="http://www.doxygen.org/index.html">
	<img class="footer" src="doxygen.png" alt="doxygen"/></a> 1.8.4 </li>
	</ul>
	</div>
	</body>
	</html>