-
Notifications
You must be signed in to change notification settings - Fork 1
/
display.readme
88 lines (79 loc) · 4.46 KB
/
display.readme
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
display.readme -- Documentation on the tree-display and animation
module of OC1.
****************************************************************
* Copyright (C) 1993, 1994 *
* Department of Computer Science *
* Johns Hopkins University *
* Sreerama K. Murthy ([email protected]) *
* Steven Salzberg ([email protected]) *
* Simon Kasif ([email protected]) *
****************************************************************
The purpose of the DISPLAY program is to display, as PostScript(R)
images, planar decision trees and data. We found this to be a useful
tool in visualizing and understanding the tree-building process.
DISPLAY can produce PostScript(R) displays for datasets alone and data
with decision trees (axis parallel or oblique). It can also produce
what we call, for lack of a better term, animations. Please note that
the animation option is still in a very experimental stage.
The following is a description of the various command line options
for "display".
-d : number of dimensions (or attributes)
If this number is not equal to 2, an error results.
It is not necessary to specify this parameter.
-D : File containing the decision tree to be displayed.
For sample format, see sample.dt.
The numbers of categories and dimensions specified in the
decision tree file override the -d and -c options.
You can also give the "animation dumps" produced by
MKTREE with the -D option.
Default: None.
-e : No erase.
When DISPLAY is used for animation, successive perturbations of
the hyperplane are animated by drawing a perturbation, waiting
for some wait_time (-w option), erasing the perturbation and
drawing the next perturbation. By using -e option, you can turn
off the erasing, so that all considered perturbations can be seen.
This can produce some really messy pictures.
-h : Character string header (title) for the display.
Default="<datafile name>" or
"<datafile name>-<decision tree file name>", depending on
which of the -t and -D options are specified.
-o : file to write the PostScript(R) output.
Default=stdout (the screen)
-t or -T : File containing the data points
For sample format, see linear.dta.
If a file name is specified with this option, the program
automatically computes the number of examples, the number of
attributes, and the number of categories.
Default: None.
-v : Verbose output. Default = FALSE
-w : wait_time
Waiting time (in arbitrary units) between displaying one
hyperplane and erasing it. Default = 100.
-x : Minimum x coordinate for the display. Default = 72. (These are
PostScript(R) screen coordinates).
-X : Maximum x coordinate for the display. Default = 540.
-y : Minimum y coordinate for the display. Default = 72.
-Y : Maximum y coordinate for the display. Default = 640.
If a datafile is specified with the -t or -T option, "display" uses
the numbers in the first column as the x coordinates, those in the
second column as the y coordinates, and the those in the third column
as the class numbers/labels. If a line has less or more than 3 columns,
an error is reported.
"display" first finds the minimum and maximum x and y coordinates
based on the dataset. These extreme coordinates, along with the
display boundary, define scaling factors along the x and y dimensions.
Animation: Notice that some PostScript(R) previewers and all printers
draw the whole file and display it, if there are no page breaks. In
such cases, the animation can not be seen in the default mode, as all
ERASES are performed before the final picture is rendered by the
postscript viewer. In this case, the -e option can be used (if the
data set is small and not very complicated) to see all hyperplane
positions considered by the program. The problem with -e option is
that the picture gets complicated and useless too fast. Viewers like
ghostscript execute postscript commands one-by-one, and hence can show
the animation.
Don't try to print postscript files produced using animation files, if
you have used any wait time. The printing may take forever.
Typically, the animation postscript files are very large even for
moderately complicated training sets.