'\" -*- coding: UTF-8 -*-
.if \n(.g .ds T< \\FC
.if \n(.g .ds T> \\F[\n[.fam]]
.de URL
\\$2 \(la\\$1\(ra\\$3
..
.if \n(.g .mso www.tmac
.TH pkregann 1 "06 December 2020" "" ""
.SH NAME
pkregann \- regression with artificial neural network (multi-layer perceptron)
.SH SYNOPSIS
'nh
.fi
.ad l
\fBpkregann\fR \kx
.if (\nx>(\n(.l/2)) .nr x (\n(.l/5)
'in \n(.iu+\nxu
\fB-i\fR \fIinput\fR \fB-t\fR \fItraining\fR [\fB-ic\fR \fIcol\fR] [\fB-oc\fR \fIcol\fR] \fB-o\fR \fIoutput\fR [\fIoptions\fR] [\fIadvanced options\fR]
'in \n(.iu-\nxu
.ad b
'hy
.SH DESCRIPTION
\fBpkregann\fR performs a regression based on an artificial
neural network.
The regression is trained from the input (\*(T<\fB\-ic\fR\*(T>) and
output (\*(T<\fB\-oc\fR\*(T>) columns in a training text file.
Each row in the training file represents one sampling unit.
Multi-dimensional input features can be defined with multiple input
options (e.g.,
\*(T<\fB\-ic\fR\*(T> \fI0\fR
\*(T<\fB\-ic\fR\*(T> \fI1\fR
\*(T<\fB\-ic\fR\*(T> \fI2\fR
for three dimensional features).
.SH OPTIONS
.TP 
\*(T<\fB\-i\fR\*(T> \fIfilename\fR, \*(T<\fB\-\-input\fR\*(T> \fIfilename\fR
input ASCII file
.TP 
\*(T<\fB\-t\fR\*(T> \fIfilename\fR, \*(T<\fB\-\-training\fR\*(T> \fIfilename\fR
training ASCII file (each row represents one sampling unit.
Input features should be provided as columns, followed by output)
.TP 
\*(T<\fB\-o\fR\*(T> \fIfilename\fR, \*(T<\fB\-\-output\fR\*(T> \fIfilename\fR
output ASCII file for result
.TP 
\*(T<\fB\-ic\fR\*(T> \fIcol\fR, \*(T<\fB\-\-inputCols\fR\*(T> \fIcol\fR
input columns (e.g., for three dimensional input data in first
three columns use:
\*(T<\fB\-ic\fR\*(T> \fI0\fR
\*(T<\fB\-ic\fR\*(T> \fI1\fR
\*(T<\fB\-ic\fR\*(T> \fI2\fR
.TP 
\*(T<\fB\-oc\fR\*(T> \fIcol\fR, \*(T<\fB\-\-outputCols\fR\*(T> \fIcol\fR
output columns (e.g., for two dimensional output in columns 3 and 4
(starting from \fI0\fR) use:
\*(T<\fB\-oc\fR\*(T> \fI3\fR
\*(T<\fB\-oc\fR\*(T> \fI4\fR
.TP 
\*(T<\fB\-from\fR\*(T> \fIrow\fR, \*(T<\fB\-\-from\fR\*(T> \fIrow\fR
start from this row in training file (start from 0)
.TP 
\*(T<\fB\-to\fR\*(T> \fIrow\fR, \*(T<\fB\-\-to\fR\*(T> \fIrow\fR
read until this row in training file (start from 0 or set leave 0
as default to read until end of file)
.TP 
\*(T<\fB\-cv\fR\*(T> \fIsize\fR, \*(T<\fB\-\-cv\fR\*(T> \fIsize\fR
n-fold cross validation mode
.TP 
\*(T<\fB\-nn\fR\*(T> \fInumber\fR, \*(T<\fB\-\-nneuron\fR\*(T> \fInumber\fR
number of neurons in hidden layers in neural network (multiple
hidden layers are set by defining multiple number of neurons:
\*(T<\fB\-n\fR\*(T> \fI15\fR
\*(T<\fB\-n\fR\*(T> \fI1\fR,
default is one hidden layer with 5 neurons)
.TP 
\*(T<\fB\-v\fR\*(T> \fIlevel\fR, \*(T<\fB\-\-verbose\fR\*(T> \fIlevel\fR
set to: 0 (results only), 1 (confusion matrix), 2 (debug)
.PP
Advanced options
.TP 
\*(T<\fB\-\-offset\fR\*(T> \fIvalue\fR
offset value for each spectral band input features:
\*(T<refl[band]=(DN[band]\-offset[band])/scale[band]\*(T>
.TP 
\*(T<\fB\-\-scale\fR\*(T> \fIvalue\fR
scale value for each spectral band input features:
\*(T<refl=(DN[band]\-offset[band])/scale[band]\*(T>
(use \*(T<0\*(T> if scale min and max in each band to
\*(T<\-1.0\*(T> and \*(T<1.0\*(T>)
.TP 
\*(T<\fB\-\-connection\fR\*(T> \fIrate\fR
connection rate (default: 1.0 for a fully connected network)
.TP 
\*(T<\fB\-l\fR\*(T> \fIrate\fR, \*(T<\fB\-\-learning\fR\*(T> \fIrate\fR
learning rate (default: 0.7)
.TP 
\*(T<\fB\-\-maxit\fR\*(T> \fInumber\fR
number of maximum iterations (epoch) (default: 500)