|
The most basic invocation of rsar is this (using the
bundled iris.dat file):
$ rsar iris.dat
reduct 0,1,2 4 1
reduct 0,2,3 4 1
This reduces the dataset iris.dat. A perfect
discernibility was attained (the last number is the value of
gamma). Two reducts were found: one involves the
dataset’s first (0), second (1) and third (2)
attributes; the other the first, third and fourth ones. The
label (or class attribute, or decision attribute) is the
fifth attribute (4). Attribute and sample numbers are
zero-based by default. This can be overridden by specifying
-1. To make the process and output less cryptic:
$ rsar -1svv iris.dat
gamma ( {1} |--> {5} ) = 0.213333
gamma ( {2} |--> {5} ) = 0.126667
gamma ( {3} |--> {5} ) = 0.806667
gamma ( {4} |--> {5} ) = 0.746667
gamma ( {1,3} |--> {5} ) = 0.986667
gamma ( {2,3} |--> {5} ) = 0.966667
gamma ( {3,4} |--> {5} ) = 0.98
gamma ( {1,2,3} |--> {5} ) = 1
A reduct was found.
Selected features: {1,2,3} |--> {5}, gamma = 1
gamma ( {1,3,4} |--> {5} ) = 1
A reduct was found.
Selected features: {1,3,4} |--> {5}, gamma = 1
Passes through the dataset: 10
Total bytes read: 21460
Total tokens (values) read: 6559
Total lines (samples) read: 1500
This shows the processing as it is carried out
(-vv, specified twice for increased verbosity),
displays attributes starting at 1 (-1) and prints
statistics at the end of processing (-s).
The first paragraph shows the results of attempting to
select the best first attribute (feature). The third
attribute is selected, as it provides a gamma of
approximately 0.8. The second paragraph shows the addition
of the second attribute. The third paragraph adds a third
attribute, and reaches two reducts, {1,2,3} and {1,3,4}.
On a POSIX operating system, it is possible to get a wide
range of statistics by specifying the -m, -s
and -t options. This particular example shows
rsar yielding all reducts of the Iris dataset on a
1.2 GHz x86-based machine with 512 Mb of main memory, and
running Linux:
$ rsar -1svAmt iris.dat
Selected features: {1,2,3} |--> {5}, gamma = 1
Selected features: {1,2,3,4} |--> {5}, gamma = 1
Selected features: {1,2,4} |--> {5}, gamma = 1
Selected features: {1,3,4} |--> {5}, gamma = 1
Selected features: {2,3,4} |--> {5}, gamma = 1
Passes through the dataset: 16
Total bytes read: 34336
Total tokens (values) read: 11054
Total lines (samples) read: 2400
User time elapsed: 0.010 seconds (1 jiffies)
System time elapsed: 0.000 seconds (0 jiffies)
Time elapsed: 0.013 seconds
Maximum memory size (RSS): 294912 bytes
Integral memory allocated: 185968 bytes
|
