<?xml version="1.0" encoding="ascii"?>
<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Transitional//EN"
"DTD/xhtml1-transitional.dtd">
<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
<head>
<title>Bio.HMM.Trainer.BaumWelchTrainer</title>
<link rel="stylesheet" href="epydoc.css" type="text/css" />
<script type="text/javascript" src="epydoc.js"></script>
</head>
<body bgcolor="white" text="black" link="blue" vlink="#204080"
alink="#204080">
<!-- ==================== NAVIGATION BAR ==================== -->
<table class="navbar" border="0" width="100%" cellpadding="0"
bgcolor="#a0c0ff" cellspacing="0">
<tr valign="middle">
<!-- Tree link -->
<th> <a
href="module-tree.html">Trees</a> </th>
<!-- Index link -->
<th> <a
href="identifier-index.html">Indices</a> </th>
<!-- Help link -->
<th> <a
href="help.html">Help</a> </th>
<th class="navbar" width="100%"></th>
</tr>
</table>
<table width="100%" cellpadding="0" cellspacing="0">
<tr valign="top">
<td width="100%">
<span class="breadcrumbs">
<a href="Bio-module.html">Package Bio</a> ::
<a href="Bio.HMM-module.html">Package HMM</a> ::
<a href="Bio.HMM.Trainer-module.html">Module Trainer</a> ::
Class BaumWelchTrainer
</span>
</td>
<td>
<table cellpadding="0" cellspacing="0">
<!-- hide/show private -->
<tr><td align="right"><span class="options">[<a href="javascript:void(0);" class="privatelink"
onclick="toggle_private();">hide private</a>]</span></td></tr>
<tr><td align="right"><span class="options"
>[<a href="frames.html" target="_top">frames</a
>] | <a href="Bio.HMM.Trainer.BaumWelchTrainer-class.html"
target="_top">no frames</a>]</span></td></tr>
</table>
</td>
</tr>
</table>
<!-- ==================== CLASS DESCRIPTION ==================== -->
<h1 class="epydoc">Class BaumWelchTrainer</h1><p class="nomargin-top"><span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer">source code</a></span></p>
<pre class="base-tree">
object --+
|
<a href="Bio.HMM.Trainer.AbstractTrainer-class.html">AbstractTrainer</a> --+
|
<strong class="uidshort">BaumWelchTrainer</strong>
</pre>
<hr />
<p>Trainer that uses the Baum-Welch algorithm to estimate parameters.</p>
<p>These should be used when a training sequence for an HMM has unknown
paths for the actual states, and you need to make an estimation of the
model parameters from the observed emissions.</p>
<p>This uses the Baum-Welch algorithm, first described in Baum, L.E.
1972. Inequalities. 3:1-8 This is based on the description in 'Biological
Sequence Analysis' by Durbin et al. in section 3.3</p>
<p>This algorithm is guaranteed to converge to a local maximum, but not
necessarily to the global maxima, so use with care!</p>
<!-- ==================== INSTANCE METHODS ==================== -->
<a name="section-InstanceMethods"></a>
<table class="summary" border="1" cellpadding="3"
cellspacing="0" width="100%" bgcolor="white">
<tr bgcolor="#70b0f0" class="table-header">
<td colspan="2" class="table-header">
<table border="0" cellpadding="0" cellspacing="0" width="100%">
<tr valign="top">
<td align="left"><span class="table-header">Instance Methods</span></td>
<td align="right" valign="top"
><span class="options">[<a href="#section-InstanceMethods"
class="privatelink" onclick="toggle_private();"
>hide private</a>]</span></td>
</tr>
</table>
</td>
</tr>
<tr>
<td width="15%" align="right" valign="top" class="summary">
<span class="summary-type"> </span>
</td><td class="summary">
<table width="100%" cellpadding="0" cellspacing="0" border="0">
<tr>
<td><span class="summary-sig"><a href="Bio.HMM.Trainer.BaumWelchTrainer-class.html#__init__" class="summary-sig-name">__init__</a>(<span class="summary-sig-arg">self</span>,
<span class="summary-sig-arg">markov_model</span>)</span><br />
Initialize the trainer.</td>
<td align="right" valign="top">
<span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer.__init__">source code</a></span>
</td>
</tr>
</table>
</td>
</tr>
<tr>
<td width="15%" align="right" valign="top" class="summary">
<span class="summary-type"> </span>
</td><td class="summary">
<table width="100%" cellpadding="0" cellspacing="0" border="0">
<tr>
<td><span class="summary-sig"><a href="Bio.HMM.Trainer.BaumWelchTrainer-class.html#train" class="summary-sig-name">train</a>(<span class="summary-sig-arg">self</span>,
<span class="summary-sig-arg">training_seqs</span>,
<span class="summary-sig-arg">stopping_criteria</span>,
<span class="summary-sig-arg">dp_method</span>=<span class="summary-sig-default"><class 'Bio.HMM.DynamicProgramming.ScaledDPAlgorithms'></span>)</span><br />
Estimate the parameters using training sequences.</td>
<td align="right" valign="top">
<span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer.train">source code</a></span>
</td>
</tr>
</table>
</td>
</tr>
<tr>
<td width="15%" align="right" valign="top" class="summary">
<span class="summary-type"> </span>
</td><td class="summary">
<table width="100%" cellpadding="0" cellspacing="0" border="0">
<tr>
<td><span class="summary-sig"><a href="Bio.HMM.Trainer.BaumWelchTrainer-class.html#update_transitions" class="summary-sig-name">update_transitions</a>(<span class="summary-sig-arg">self</span>,
<span class="summary-sig-arg">transition_counts</span>,
<span class="summary-sig-arg">training_seq</span>,
<span class="summary-sig-arg">forward_vars</span>,
<span class="summary-sig-arg">backward_vars</span>,
<span class="summary-sig-arg">training_seq_prob</span>)</span><br />
Add the contribution of a new training sequence to the transitions.</td>
<td align="right" valign="top">
<span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer.update_transitions">source code</a></span>
</td>
</tr>
</table>
</td>
</tr>
<tr>
<td width="15%" align="right" valign="top" class="summary">
<span class="summary-type"> </span>
</td><td class="summary">
<table width="100%" cellpadding="0" cellspacing="0" border="0">
<tr>
<td><span class="summary-sig"><a href="Bio.HMM.Trainer.BaumWelchTrainer-class.html#update_emissions" class="summary-sig-name">update_emissions</a>(<span class="summary-sig-arg">self</span>,
<span class="summary-sig-arg">emission_counts</span>,
<span class="summary-sig-arg">training_seq</span>,
<span class="summary-sig-arg">forward_vars</span>,
<span class="summary-sig-arg">backward_vars</span>,
<span class="summary-sig-arg">training_seq_prob</span>)</span><br />
Add the contribution of a new training sequence to the emissions</td>
<td align="right" valign="top">
<span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer.update_emissions">source code</a></span>
</td>
</tr>
</table>
</td>
</tr>
<tr>
<td colspan="2" class="summary">
<p class="indent-wrapped-lines"><b>Inherited from <code><a href="Bio.HMM.Trainer.AbstractTrainer-class.html">AbstractTrainer</a></code></b>:
<code><a href="Bio.HMM.Trainer.AbstractTrainer-class.html#estimate_params">estimate_params</a></code>,
<code><a href="Bio.HMM.Trainer.AbstractTrainer-class.html#log_likelihood">log_likelihood</a></code>,
<code><a href="Bio.HMM.Trainer.AbstractTrainer-class.html#ml_estimator">ml_estimator</a></code>
</p>
<p class="indent-wrapped-lines"><b>Inherited from <code>object</code></b>:
<code>__delattr__</code>,
<code>__format__</code>,
<code>__getattribute__</code>,
<code>__hash__</code>,
<code>__new__</code>,
<code>__reduce__</code>,
<code>__reduce_ex__</code>,
<code>__repr__</code>,
<code>__setattr__</code>,
<code>__sizeof__</code>,
<code>__str__</code>,
<code>__subclasshook__</code>
</p>
</td>
</tr>
</table>
<!-- ==================== PROPERTIES ==================== -->
<a name="section-Properties"></a>
<table class="summary" border="1" cellpadding="3"
cellspacing="0" width="100%" bgcolor="white">
<tr bgcolor="#70b0f0" class="table-header">
<td colspan="2" class="table-header">
<table border="0" cellpadding="0" cellspacing="0" width="100%">
<tr valign="top">
<td align="left"><span class="table-header">Properties</span></td>
<td align="right" valign="top"
><span class="options">[<a href="#section-Properties"
class="privatelink" onclick="toggle_private();"
>hide private</a>]</span></td>
</tr>
</table>
</td>
</tr>
<tr>
<td colspan="2" class="summary">
<p class="indent-wrapped-lines"><b>Inherited from <code>object</code></b>:
<code>__class__</code>
</p>
</td>
</tr>
</table>
<!-- ==================== METHOD DETAILS ==================== -->
<a name="section-MethodDetails"></a>
<table class="details" border="1" cellpadding="3"
cellspacing="0" width="100%" bgcolor="white">
<tr bgcolor="#70b0f0" class="table-header">
<td colspan="2" class="table-header">
<table border="0" cellpadding="0" cellspacing="0" width="100%">
<tr valign="top">
<td align="left"><span class="table-header">Method Details</span></td>
<td align="right" valign="top"
><span class="options">[<a href="#section-MethodDetails"
class="privatelink" onclick="toggle_private();"
>hide private</a>]</span></td>
</tr>
</table>
</td>
</tr>
</table>
<a name="__init__"></a>
<div>
<table class="details" border="1" cellpadding="3"
cellspacing="0" width="100%" bgcolor="white">
<tr><td>
<table width="100%" cellpadding="0" cellspacing="0" border="0">
<tr valign="top"><td>
<h3 class="epydoc"><span class="sig"><span class="sig-name">__init__</span>(<span class="sig-arg">self</span>,
<span class="sig-arg">markov_model</span>)</span>
<br /><em class="fname">(Constructor)</em>
</h3>
</td><td align="right" valign="top"
><span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer.__init__">source code</a></span>
</td>
</tr></table>
<p>Initialize the trainer.</p>
<p>Arguments:</p>
<p>o markov_model - The model we are going to estimate parameters for.
This should have the parameters with some initial estimates, that we can
build from.</p>
<dl class="fields">
<dt>Overrides:
object.__init__
</dt>
</dl>
</td></tr></table>
</div>
<a name="train"></a>
<div>
<table class="details" border="1" cellpadding="3"
cellspacing="0" width="100%" bgcolor="white">
<tr><td>
<table width="100%" cellpadding="0" cellspacing="0" border="0">
<tr valign="top"><td>
<h3 class="epydoc"><span class="sig"><span class="sig-name">train</span>(<span class="sig-arg">self</span>,
<span class="sig-arg">training_seqs</span>,
<span class="sig-arg">stopping_criteria</span>,
<span class="sig-arg">dp_method</span>=<span class="sig-default"><class 'Bio.HMM.DynamicProgramming.ScaledDPAlgorithms'></span>)</span>
</h3>
</td><td align="right" valign="top"
><span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer.train">source code</a></span>
</td>
</tr></table>
<p>Estimate the parameters using training sequences.</p>
<p>The algorithm for this is taken from Durbin et al. p64, so this is a
good place to go for a reference on what is going on.</p>
<p>Arguments:</p>
<p>o training_seqs -- A list of TrainingSequence objects to be used for
estimating the parameters.</p>
<p>o stopping_criteria -- A function, that when passed the change in log
likelihood and threshold, will indicate if we should stop the estimation
iterations.</p>
<p>o dp_method -- A class instance specifying the dynamic programming
implementation we should use to calculate the forward and backward
variables. By default, we use the scaling method.</p>
<dl class="fields">
</dl>
</td></tr></table>
</div>
<a name="update_transitions"></a>
<div>
<table class="details" border="1" cellpadding="3"
cellspacing="0" width="100%" bgcolor="white">
<tr><td>
<table width="100%" cellpadding="0" cellspacing="0" border="0">
<tr valign="top"><td>
<h3 class="epydoc"><span class="sig"><span class="sig-name">update_transitions</span>(<span class="sig-arg">self</span>,
<span class="sig-arg">transition_counts</span>,
<span class="sig-arg">training_seq</span>,
<span class="sig-arg">forward_vars</span>,
<span class="sig-arg">backward_vars</span>,
<span class="sig-arg">training_seq_prob</span>)</span>
</h3>
</td><td align="right" valign="top"
><span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer.update_transitions">source code</a></span>
</td>
</tr></table>
<p>Add the contribution of a new training sequence to the
transitions.</p>
<p>Arguments:</p>
<p>o transition_counts -- A dictionary of the current counts for the
transitions</p>
<p>o training_seq -- The training sequence we are working with</p>
<p>o forward_vars -- Probabilities calculated using the forward
algorithm.</p>
<p>o backward_vars -- Probabilities calculated using the backwards
algorithm.</p>
<p>o training_seq_prob - The probability of the current sequence.</p>
<p>This calculates A_{kl} (the estimated transition counts from state k
to state l) using formula 3.20 in Durbin et al.</p>
<dl class="fields">
</dl>
</td></tr></table>
</div>
<a name="update_emissions"></a>
<div>
<table class="details" border="1" cellpadding="3"
cellspacing="0" width="100%" bgcolor="white">
<tr><td>
<table width="100%" cellpadding="0" cellspacing="0" border="0">
<tr valign="top"><td>
<h3 class="epydoc"><span class="sig"><span class="sig-name">update_emissions</span>(<span class="sig-arg">self</span>,
<span class="sig-arg">emission_counts</span>,
<span class="sig-arg">training_seq</span>,
<span class="sig-arg">forward_vars</span>,
<span class="sig-arg">backward_vars</span>,
<span class="sig-arg">training_seq_prob</span>)</span>
</h3>
</td><td align="right" valign="top"
><span class="codelink"><a href="Bio.HMM.Trainer-pysrc.html#BaumWelchTrainer.update_emissions">source code</a></span>
</td>
</tr></table>
<p>Add the contribution of a new training sequence to the emissions</p>
<p>Arguments:</p>
<p>o emission_counts -- A dictionary of the current counts for the
emissions</p>
<p>o training_seq -- The training sequence we are working with</p>
<p>o forward_vars -- Probabilities calculated using the forward
algorithm.</p>
<p>o backward_vars -- Probabilities calculated using the backwards
algorithm.</p>
<p>o training_seq_prob - The probability of the current sequence.</p>
<p>This calculates E_{k}(b) (the estimated emission probability for
emission letter b from state k) using formula 3.21 in Durbin et al.</p>
<dl class="fields">
</dl>
</td></tr></table>
</div>
<br />
<!-- ==================== NAVIGATION BAR ==================== -->
<table class="navbar" border="0" width="100%" cellpadding="0"
bgcolor="#a0c0ff" cellspacing="0">
<tr valign="middle">
<!-- Tree link -->
<th> <a
href="module-tree.html">Trees</a> </th>
<!-- Index link -->
<th> <a
href="identifier-index.html">Indices</a> </th>
<!-- Help link -->
<th> <a
href="help.html">Help</a> </th>
<th class="navbar" width="100%"></th>
</tr>
</table>
<table border="0" cellpadding="0" cellspacing="0" width="100%%">
<tr>
<td align="left" class="footer">
Generated by Epydoc 3.0.1 on Thu Aug 18 18:19:26 2011
</td>
<td align="right" class="footer">
<a target="mainFrame" href="http://epydoc.sourceforge.net"
>http://epydoc.sourceforge.net</a>
</td>
</tr>
</table>
<script type="text/javascript">
<!--
// Private objects are initially displayed (because if
// javascript is turned off then we want them to be
// visible); but by default, we want to hide them. So hide
// them unless we have a cookie that says to show them.
checkCookie();
// -->
</script>
</body>
</html>
