NAME
analyze - analyzes result files generated by the neural
network simulation package SNNS
SYNOPSIS
analyze [-w] [-r] [-u] [-a] [-S " t c ] [-v]
[-s] [-c] [-m] [ -i filename] [ -o filename] [-e
function ] [ -l lower-bound ] [-h upper-bound ]
OPTIONS
-w
print number of patterns which were classified wrong.
(default)
-r
print number of patterns which were classified
correctly.
-u
print number of patterns which were not classified.
-a
same as '-w -r -u'
-S " t c "
specific: numbers of patterns of class t which are
classified as class c are printed (-1 = noclass)
-v
verbose output. Each printed number is preceded by one of
the words "wrong", "right", "unknown", or
"specific" depending on the result of the classification.
-s
statistic information containing wrong, right and not
classified patterns. The network error is printed also.
-c
same as -s, but statistics for each output unit
(class) is displayed.
-m
show confusion matrix (only works with -e 402040 or -e
WTA)
-i filename
name of the 'result file' which is going to be analyzed.
Without this option file is read from stdin.
-o filename
name of the file which is going to be produced by
analyze. Default is stdout.
-e function
defines the name of the 'analyzing function' to be used.
function can be one of the following:
402040
The '402040' rule. That means on a range from 0 to 1
h will be 0.6 (upper 40%) and l will be 0.4 (lower 40%). The
middle 20% is represented by h - l. The classification of the patterns
will depend on the values of h, l, and other constrains.
WTA
Winner takes all. That means the classification depends
on the unit with the highest output, on the values of h, l, and
other constrains.
band
An analyzing function that checks a band of values around
the teaching output. The classification of the patterns will depend on the
values of h, l, and other constrains.
-l lower-bound
first parameter of the analyzing function.
-h upper-bound
second parameter of the analyzing function.
DESCRIPTION
The purpose of this tool is to analyze the result files that have
been created by SNNS. The result file which you want to analyze has to
contain the teaching output and the propagated output of the
network. This tool will let you quantify the performance of your network
much better than the mere SSE value obtained during training.
SEE
ALSO
The SNNS user manual under the SNNS installation directory or the
hypertext manual at
http://www-ra.informatik.uni-tuebingen.de/SNNS/UserManual/UserManual.html
snns(1) description of the whole package
