galaxy-central / tools / phenotype_association / lps.xml

Full commit
<tool id="hgv_lps" name="LPS" version="1.0.0">
  <description>LASSO-Patternsearch algorithm</description>

  <command interpreter="bash"> $lambda_fac $input_file $label_column $output_file $log_file
    Initialization 0
    #if $advanced.options == "true":
      Sample $advanced.sample
      Verbosity $advanced.verbosity
      Standardize $advanced.standardize
      initialLambda $advanced.initialLambda
      #if $advanced.continuation.continuation == "1":
        Continuation $advanced.continuation.continuation
        continuationSteps $advanced.continuation.continuationSteps
        accurateIntermediates $advanced.continuation.accurateIntermediates
      #end if
      printFreq $advanced.printFreq
      #if $advanced.newton.newton == "1":
        Newton $advanced.newton.newton
        NewtonThreshold $advanced.newton.newtonThreshold
      #end if
      HessianSampleFraction $advanced.hessianSampleFraction
      BB 0
      Monotone 0
      FullGradient $advanced.fullGradient
      GradientFraction $advanced.gradientFraction
      InitialAlpha $advanced.initialAlpha
      AlphaIncrease $advanced.alphaIncrease
      AlphaDecrease $advanced.alphaDecrease
      AlphaMax $advanced.alphaMax
      c1 $advanced.c1
      MaxIter $advanced.maxIter
      StopTol $advanced.stopTol
      IntermediateTol $advanced.intermediateTol
      FinalOnly $advanced.finalOnly
    #end if

    <param name="input_file" type="data" format="tabular" label="Dataset"/>
    <param name="label_column" type="data_column" data_ref="input_file" numerical="true" label="Label column" help="Column containing outcome labels: +1 or -1."/>
    <param name="lambda_fac" label="Lambda_fac" type="float" value="0.03" help="Target value of the regularization parameter, expressed as a fraction of the calculated lambda_max.">
      <validator type="in_range" message="0.00 &lt; lambda_fac &lt;= 1.00" min="0.00" max="1.00"/>
    <conditional name="advanced">
      <param name="options" type="select" label="Advanced Options">
        <option value="false" selected="true">Hide advanced options</option>
        <option value="true">Show advanced options</option>
      <when value="false">
        <!-- no options -->
      <when value="true">
        <!-- HARDCODED: 'Sample' we don't support passing an array -->
        <param name="sample" type="float" value="1.0" label="Sample fraction" help="Sample this fraction of the data set.">
          <validator type="in_range" message="0.0 &lt;= sample &lt;= 1.0" min="0.0" max="1.0"/>
        <!-- HARDCODED: 'Initialization' = 0 :: Initialize at beta=0 -->
        <param name="verbosity" type="select" format="integer" label="Verbosity">
          <option value="0" selected="true">Little output</option>
          <option value="1">More output</option>
          <option value="2">Still more output</option>
        <param name="standardize" type="select" format="integer" label="Standardize" help="Scales and shifts each column so that it has mean zero and variance 1.">
          <option value="0" selected="true">Don't standardize</option>
          <option value="1">Standardize</option>
        <param name="initialLambda" type="float" value="0.8" label="Initial lambda" help="First value of lambda to be used in the continuation scheme, expressed as a fraction of lambda_max.">
          <validator type="in_range" message="0.0 &lt; initialLambda &lt; 1.0" min="0.0" max="1.0"/>
        <conditional name="continuation">
          <param name="continuation" type="select" format="integer" label="Continuation" help="Use continuation strategy to start with a larger value of lambda, decreasing it successively to lambda_fac.">
            <option value="0" selected="true">Don't use continuation</option>
            <option value="1">Use continuation</option>
          <when value="0">
            <!-- no options -->
          <when value="1">
            <param name="continuationSteps" type="integer" value="5" label="Continuation steps" help="Number of lambda values to use in continuation &lt;em&gt;prior&lt;/em&gt; to target value lambda_fac."/>

            <param name="accurateIntermediates" type="select" format="integer" label="Accurate intermediates" help="Indicates whether accurate solutions are required for lambda values other than the target value lambda_fac.">
              <option value="0" selected="true">Don't need accurate intemediates</option>
              <option value="1">Calculate accurate intermediates</option>
        </conditional> <!-- name="continuation" -->
        <param name="printFreq" type="integer" value="1" label="Print frequency" help="Print a progress report every NI iterations, where NI is the supplied value of this parameter.">
          <validator type="in_range" message="printFreq &gt;= 1" min="1"/>
        <conditional name="newton">
          <param name="newton" type="select" format="integer" label="Projected Newton steps">
            <option value="0" selected="true">No Newton steps</option>
            <option value="1">Try projected Newton steps</option>
          <when value="0">
            <!-- no options -->
          <when value="1">
            <param name="newtonThreshold" type="integer" value="500" label="Newton threshold" help="Maximum size of free variable subvector for Newton."/>
        <param name="hessianSampleFraction" type="float" value="1.0" label="Hessian sample fraction" help="Fraction of terms to use in approximate Hessian calculation.">
          <validator type="in_range" message="0.01 &lt; hessianSampleFraction &lt;= 1.00" min="0.01" max="1.00"/>
        <!-- HARDCODED: 'BB' = 0 :: don't use Barzilai-Borwein steps -->
        <!-- HARDCODED: 'Monotone' = 0 :: don't force monotonicity -->
        <param name="fullGradient" type="select" format="integer" label="Partial gradient vector selection">
          <option value="0">Use randomly selected partial gradient, including current active components ("biased")</option>
          <option value="1">Use full gradient vector at every step</option>
          <option value="2">Randomly selected partial gradient, without regard to current active set ("unbiased")</option>
        <param name="gradientFraction" type="float" value="0.1" label="Gradient fraction" help="Fraction of inactive gradient vector to evaluate.">
          <validator type="in_range" message="0.0 &lt; gradientFraction &lt;= 1" min="0.0" max="1.0"/>
        <param name="initialAlpha" type="float" value="1.0" label="Initial value of alpha"/>
        <param name="alphaIncrease" type="float" value="2.0" label="Alpha increase" help="Factor by which to increase alpha after descent not obtained."/>
        <param name="alphaDecrease" type="float" value="0.8" label="Alpha decrease" help="Factor by which to decrease alpha after successful first-order step."/>
        <param name="alphaMax" type="float" value="1e12" label="Alpha max" help="Maximum value of alpha; terminate with error if we exceed this."/>
        <param name="c1" type="float" value="1e-3" help="Parameter defining the margin by which the first-order step is required to decrease before being taken.">
          <validator type="in_range" message="0.0 &lt; c1 &lt; 1.0" min="0.0" max="1.0"/>
        <param name="maxIter" type="integer" value="10000" label="Maximum number of iterations" help="Terminate with error if we exceed this."/>
        <param name="stopTol" type="float" value="1e-6" label="Stop tolerance" help="Convergence tolerance for target value of lambda."/>
        <param name="intermediateTol" type="float" value="1e-4" label="Intermediate tolerance" help="Convergence tolerance for intermediate values of lambda."/>
        <param name="finalOnly" type="select" format="integer" label="Final only">
          <option value="0" selected="true">Return information for all intermediate values</option>
          <option value="1">Just return information at the last lambda</option>
      </when> <!-- value="advanced" -->
    </conditional> <!-- name="advanced" -->

    <data name="output_file" format="tabular" label="${} on ${on_string}: results"/>
    <data name="log_file" format="txt" label="${} on ${on_string}: log"/>

    <requirement type="package">lps_tool</requirement>

      <param name="input_file" value="lps_arrhythmia.tabular"/>
      <param name="label_column" value="280"/>
      <param name="lambda_fac" value="0.03"/>
      <param name="options" value="true"/>
      <param name="sample" value="1.0"/>
      <param name="verbosity" value="1"/>
      <param name="standardize" value="0"/>
      <param name="initialLambda" value="0.9"/>
      <param name="continuation" value="1"/>
      <param name="continuationSteps" value="10"/>
      <param name="accurateIntermediates" value="0"/>
      <param name="printFreq" value="1"/>
      <param name="newton" value="1"/>
      <param name="newtonThreshold" value="500"/>
      <param name="hessianSampleFraction" value="1.0"/>
      <param name="fullGradient" value="1"/>
      <param name="gradientFraction" value="0.5"/>
      <param name="initialAlpha" value="1.0"/>
      <param name="alphaIncrease" value="2.0"/>
      <param name="alphaDecrease" value="0.8"/>
      <param name="alphaMax" value="1e12"/>
      <param name="c1" value="1e-3"/>
      <param name="maxIter" value="2500"/>
      <param name="stopTol" value="1e-6"/>
      <param name="intermediateTol" value="1e-6"/>
      <param name="finalOnly" value="0"/>
      <output name="ouput_file" file="lps_arrhythmia_beta.tabular"/>
      <output name="log_file" file="lps_arrhythmia_log.txt"/>

**Dataset formats**

The input and output datasets are tabular_.  The columns are described below.
There is a second output dataset (a log) that is in text_ format.
(`Dataset missing?`_)

.. _tabular: ./static/formatHelp.html#tab
.. _text: ./static/formatHelp.html#text
.. _Dataset missing?: ./static/formatHelp.html


**What it does**

The LASSO-Patternsearch algorithm fits your dataset to an L1-regularized
logistic regression model.  A benefit of using L1-regularization is
that it typically yields a weight vector with relatively few non-zero

For example, say you have a dataset containing M rows (subjects)
and N columns (attributes) where one of these N attributes is binary,
indicating whether or not the subject has some property of interest P.
In simple terms, LPS calculates a weight for each of the other attributes
in your dataset.  This weight indicates how "relevant" that attribute
is for predicting whether or not a given subject has property P.
The L1-regularization causes most of these weights to be equal to zero,
which means LPS will find a "small" subset of the remaining N-1 attributes
in your dataset that can be used to predict P.

In other words, LPS can be used for feature selection.

The input dataset is tabular, and must contain a label column which
indicates whether or not a given row has property P.  In the current
version of this tool, P must be encoded using +1 and -1.  The Lambda_fac
parameter ranges from 0 to 1, and controls how sparse the weight
vector will be.  At the low end, when Lambda_fac = 0, there will be
no regularization.  At the high end, when Lambda_fac = 1, there will be
"too much" regularization, and all of the weights will equal zero.

The LPS tool creates two output datasets.  The first, called the results
file, is a tabular dataset containing one column of weights for each
value of the regularization parameter lambda that was tried.  The weight
columns are in order from left to right by decreasing values of lambda.
The first N-1 rows in each column are the weights for the N-1 attributes
in your input dataset.  The final row is a constant, the intercept.

Let **x** be a row from your input dataset and let **b** be a column
from the results file.  To compute the probability that row **x** has
a label value of +1:

  Probability(row **x** has label value = +1) = 1 / [1 + exp{**x** \* **b**\[1..N-1\] + **b**\[N\]}]

where **x** \* **b**\[1..N-1\] represents matrix multiplication.

The second output dataset, called the log file, is a text file which
contains additional data about the fitted L1-regularized logistic
regression model.  These data include the number of features, the
computed value of lambda_max, the actual values of lambda used, the
optimal values of the log-likelihood and regularized log-likelihood
functions, the number of non-zeros, and the number of iterations.




- input file::

    +1   1   0   0   0   0   1   0   1   1   ...
    +1   1   1   1   0   0   1   0   1   1   ...
    +1   1   0   1   0   1   0   1   0   1   ...

- output results file::


- output log file::

    Data set has 100 vectors with 50 features.
      calculateLambdaMax: n=50, m=100, m+=50, m-=50
      computed value of lambda_max: 5.0000e-01
    lambda=2.96e-02 solution:
      optimal log-likelihood function value: 6.46e-01
      optimal *regularized* log-likelihood function value: 6.79e-01
      number of nonzeros at the optimum:      5
      number of iterations required:     43



Koh K, Kim S-J, Boyd S. (2007)
An interior-point method for large-scale l1-regularized logistic regression.
Journal of Machine Learning Research. 8:1519-1555.

Shi W, Wahba G, Wright S, Lee K, Klein R, Klein B. (2008)
LASSO-Patternsearch algorithm with application to ophthalmology and genomic data.
Stat Interface. 1(1):137-153.

Wright S, Novak R, Figueiredo M. (2009)
Sparse reconstruction via separable approximation.
IEEE Transactions on Signal Processing. 57:2479-2403.

Shi J, Yin W, Osher S, Sajda P. (2010)
A fast hybrid algorithm for large scale l1-regularized logistic regression.
Journal of Machine Learning Research. 11:713-741.

Byrd R, Chin G, Neveitt W, Nocedal J. (2010)
On the use of stochastic Hessian information in unconstrained optimization.
Technical Report. Northwestern University. June 16, 2010.

Wright S. (2010)
Accelerated block-coordinate relaxation for regularized optimization.
Technical Report. University of Wisconsin. August 10, 2010.