home
***
CD-ROM
|
disk
|
FTP
|
other
***
search
/
CD ROM Paradise Collection 4
/
CD ROM Paradise Collection 4 1995 Nov.iso
/
science
/
neumap3.zip
/
NUMP.ZP
/
SOMC.HLP
< prev
next >
Wrap
Text File
|
1993-01-04
|
2KB
|
48 lines
1. Input Data Formats;
a. Each pattern must have inputs followed by 0 or more
outputs. Therefore, training data files will work.
b. Training data for classification typically has N features
followed by the class id.
c. Training data for mapping typically has N
features followed by several desired output values.
2. Output Data Format;
Output files from clustering include the number
of clusters, followed by the cluster vectors themselves.
3. Self-Organizing Map;
a. Cluster a data file using Kohonen's Self-Organizing
Feature Map.
b. Desired outputs, if any, can be ignored.
4. Processing Example for Self-Organizing Map
a. Under the "Neural Clustering" option, choose "Batch Processing"
b. From the parameter file,
8 ! number of inputs in a pattern
7 ! number of outputs in a pattern (class id not used)
Twod.tra ! filename for shape recognition training set
0 ! read all patterns in the file
1 ! initialize clusters
36 ! pick 36 as the number of clusters
20 ! number of iterations
2 ! use linearly decreasing learning factor and neighborhoods
.8 5 ! initial learning factor and half-neighborhood size
1 ! continue clustering
5 ! number of iterations
2 ! use linearly decreasing learning factor and neighborhoods
.04 0 ! initial learning factor and half-neighborhood size
2 ! stop
1 ! save clusters
sm ! filename for saved clusters
we see that the program will apply Self-Organizing Map clustering
to the file Twod.tra with 20 iterations. The number of random initial
clusters is 36. The initial learning factor and half-neighborhood
size are respectively .8 and 5, and linearly decreasing neighborhoods
and learning factor are chosen. After 20 iterations, 5 additional
iterations are specified. The clusters will be saved in a file
called sm.
c. After running the program, we can "Examine Program Output",
where we observe that the normalized clustering error is 2.84466.