[pktools] 05/07: Add man page for pkfsann.
Bas Couwenberg
sebastic at xs4all.nl
Sun Dec 7 01:42:20 UTC 2014
This is an automated email from the git hooks/post-receive script.
sebastic-guest pushed a commit to branch master
in repository pktools.
commit 0045cdea0417577a15ca52e08d61df850ce8490f
Author: Bas Couwenberg <sebastic at xs4all.nl>
Date: Sun Dec 7 02:04:36 2014 +0100
Add man page for pkfsann.
---
debian/changelog | 2 +-
debian/man/pkfsann.1.xml | 346 +++++++++++++++++++++++++++++++++++++++++++++++
2 files changed, 347 insertions(+), 1 deletion(-)
diff --git a/debian/changelog b/debian/changelog
index 7de4617..534cb50 100644
--- a/debian/changelog
+++ b/debian/changelog
@@ -5,7 +5,7 @@ pktools (2.6.1-1) UNRELEASED; urgency=medium
* Remove libbase package, library no longer installed.
* Add man page for pkann, pkascii2img, pkascii2ogr, pkcomposite, pkcreatect,
pkcrop, pkdiff, pkdsm2shadow, pkdumpimg, pkdumpogr, pkegcs, pkextract,
- pkfillnodata, pkfilter, pkfilterascii, pkfilterdem.
+ pkfillnodata, pkfilter, pkfilterascii, pkfilterdem, pkfsann.
-- Bas Couwenberg <sebastic at xs4all.nl> Wed, 03 Dec 2014 21:16:31 +0100
diff --git a/debian/man/pkfsann.1.xml b/debian/man/pkfsann.1.xml
new file mode 100644
index 0000000..a6aee70
--- /dev/null
+++ b/debian/man/pkfsann.1.xml
@@ -0,0 +1,346 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE refentry PUBLIC "-//OASIS//DTD DocBook XML V4.4//EN" "http://www.oasis-open.org/docbook/xml/4.4/docbookx.dtd">
+<refentry id='pkfsann'>
+
+ <refmeta>
+ <refentrytitle>pkfsann</refentrytitle>
+ <manvolnum>1</manvolnum>
+ </refmeta>
+
+ <refnamediv>
+ <refname>pkfsann</refname>
+ <refpurpose>feature selection for nn classifier</refpurpose>
+ </refnamediv>
+
+ <refsynopsisdiv id='synopsis'>
+ <cmdsynopsis>
+ <command>pkfsann</command>
+ <arg choice='plain'><option>-t</option> <replaceable>training</replaceable></arg>
+ <arg choice='plain'><option>-n</option> <replaceable>number</replaceable></arg>
+ <arg choice='opt'><replaceable>options</replaceable></arg>
+ <arg choice='opt'><replaceable>advanced options</replaceable></arg>
+ </cmdsynopsis>
+ </refsynopsisdiv>
+
+ <refsect1 id='description'>
+ <title>DESCRIPTION</title>
+ <para>
+ Classification problems dealing with high dimensional input data can be
+ challenging due to the Hughes phenomenon.
+ Hyperspectral data, for instance, can have hundreds of spectral bands and
+ require special attention when being classified.
+ In particular when limited training data are available,
+ the classification of such data can be problematic without reducing the
+ dimension.
+ </para>
+ <para>
+ <command>pkfsann</command> implements a number of feature selection
+ techniques, among which a sequential floating forward search (SFFS).
+ Also consider the SVM classifier implemented in
+ <citerefentry>
+ <refentrytitle>pksvm</refentrytitle>
+ <manvolnum>1</manvolnum>
+ </citerefentry>,
+ which has been shown to be more robust to this type of problem than others.
+ </para>
+ </refsect1>
+
+ <refsect1 id='options'>
+ <title>OPTIONS</title>
+ <variablelist>
+
+ <varlistentry>
+ <term><option>-t</option> <replaceable>filename</replaceable></term>
+ <term><option>--training</option> <replaceable>filename</replaceable></term>
+ <listitem>
+ <para>
+ training vector file.
+ A single vector file contains all training features
+ (must be set as: B0, B1, B2,...) for all classes
+ (class numbers identified by label option).
+ Use multiple training files for bootstrap aggregation
+ (alternative to the bag and bsize options,
+ where a random subset is taken from a single training file)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-n</option> <replaceable>number</replaceable></term>
+ <term><option>--nf</option> <replaceable>number</replaceable></term>
+ <listitem>
+ <para>
+ number of features to select
+ (0 to select optimal number,
+ see also <option>--ecost</option> option)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-i</option> <replaceable>filename</replaceable></term>
+ <term><option>--input</option> <replaceable>filename</replaceable></term>
+ <listitem>
+ <para>
+ input test set (leave empty to perform a cross validation based on
+ training only)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-v</option> <replaceable>level</replaceable></term>
+ <term><option>--verbose</option> <replaceable>level</replaceable></term>
+ <listitem>
+ <para>
+ set to: 0 (results only), 1 (confusion matrix), 2 (debug)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ </variablelist>
+
+ <para>Advanced options</para>
+ <variablelist>
+
+ <varlistentry>
+ <term><option>-tln</option> <replaceable>layer</replaceable></term>
+ <term><option>--tln</option> <replaceable>layer</replaceable></term>
+ <listitem>
+ <para>
+ training layer name(s)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-label</option> <replaceable>attribute</replaceable></term>
+ <term><option>--label</option> <replaceable>attribute</replaceable></term>
+ <listitem>
+ <para>
+ identifier for class label in training vector file.
+ (default: label)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>--balance</option> <replaceable>size</replaceable></term>
+ <listitem>
+ <para>
+ balance the input data to this number of samples for each class
+ (default: 0)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-random</option></term>
+ <term><option>--random</option></term>
+ <listitem>
+ <para>
+ in case of balance, randomize input data
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-min</option> <replaceable>number</replaceable></term>
+ <term><option>--min</option> <replaceable>number</replaceable></term>
+ <listitem>
+ <para>
+ if number of training pixels is less then min,
+ do not take this class into account
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-b</option> <replaceable>band</replaceable></term>
+ <term><option>--band</option> <replaceable>band</replaceable></term>
+ <listitem>
+ <para>
+ band index (starting from 0, either use band option or use start
+ to end)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-s</option> <replaceable>band</replaceable></term>
+ <term><option>--start</option> <replaceable>band</replaceable></term>
+ <listitem>
+ <para>
+ start band sequence number
+ (default: 0)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-e</option> <replaceable>band</replaceable></term>
+ <term><option>--end</option> <replaceable>band</replaceable></term>
+ <listitem>
+ <para>
+ end band sequence number (set to 0 to include bands)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>--offset</option> <replaceable>value</replaceable></term>
+ <listitem>
+ <para>
+ offset value for each spectral band input features:
+ refl[band]=(DN[band]-offset[band])/scale[band]
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>--scale</option> <replaceable>value</replaceable></term>
+ <listitem>
+ <para>
+ scale value for each spectral band input features:
+ refl=(DN[band]-offset[band])/scaleband
+ (use 0 if scale min and max in each band to -1.0 and 1.0)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-a</option> <replaceable>0|1|2</replaceable></term>
+ <term><option>--aggreg</option> <replaceable>0|1|2</replaceable></term>
+ <listitem>
+ <para>
+ how to combine aggregated classifiers, see also
+ <option>--rc</option> option
+ (0: no aggregation, 1: sum rule, 2: max rule).
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-sm</option> <replaceable>method</replaceable></term>
+ <term><option>--sm</option> <replaceable>method</replaceable></term>
+ <listitem>
+ <para>
+ feature selection method
+ (sffs=sequential floating forward search,
+ sfs=sequential forward search, sbs, sequential backward search,
+ bfs=brute force search)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-ecost</option> <replaceable>value</replaceable></term>
+ <term><option>--ecost</option> <replaceable>value</replaceable></term>
+ <listitem>
+ <para>
+ epsilon for stopping criterion in cost function to determine
+ optimal number of features
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-cv</option> <replaceable>value</replaceable></term>
+ <term><option>--cv</option> <replaceable>value</replaceable></term>
+ <listitem>
+ <para>
+ n-fold cross validation mode (default: 0)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-c</option> <replaceable>name</replaceable></term>
+ <term><option>--class</option> <replaceable>name</replaceable></term>
+ <listitem>
+ <para>
+ list of class names.
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-r</option> <replaceable>value</replaceable></term>
+ <term><option>--reclass</option> <replaceable>value</replaceable></term>
+ <listitem>
+ <para>
+ list of class values (use same order as in
+ <option>--class</option> option).
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-n</option> <replaceable>number</replaceable></term>
+ <term><option>--nneuron</option> <replaceable>number</replaceable></term>
+ <listitem>
+ <para>
+ number of neurons in hidden layers in neural network (multiple
+ hidden layers are set by defining multiple number of neurons:
+ <option>-nn</option> 15 <option>-nn</option> 1, default is one
+ hidden layer with 5 neurons)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>--connection</option> <replaceable>0|1</replaceable></term>
+ <listitem>
+ <para>
+ connection rate (default: 1.0 for a fully connected network)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-w</option> <replaceable>weights</replaceable></term>
+ <term><option>--weights</option> <replaceable>weights</replaceable></term>
+ <listitem>
+ <para>
+ weights for neural network.
+ Apply to fully connected network only, starting from first input
+ neuron to last output neuron, including the bias neurons (last
+ neuron in each but last layer)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>-l</option> <replaceable>rate</replaceable></term>
+ <term><option>--learning</option> <replaceable>rate</replaceable></term>
+ <listitem>
+ <para>
+ learning rate (default: 0.7)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ <varlistentry>
+ <term><option>--maxit</option> <replaceable>number</replaceable></term>
+ <listitem>
+ <para>
+ number of maximum iterations (epoch) (default: 500)
+ </para>
+ </listitem>
+ </varlistentry>
+
+ </variablelist>
+
+ </refsect1>
+
+ <refsect1 id='see-also'>
+ <title>SEE ALSO</title>
+
+ <citerefentry>
+ <refentrytitle>pksvm</refentrytitle>
+ <manvolnum>1</manvolnum>
+ </citerefentry>
+
+ </refsect1>
+
+</refentry>
--
Alioth's /usr/local/bin/git-commit-notice on /srv/git.debian.org/git/pkg-grass/pktools.git
More information about the Pkg-grass-devel
mailing list