Train SOM

SOM Analyst

Train SOM

Trains a SOM.

ArcGIS Reference

../../_images/stage1.png

Parameters

initial SOM
The initial SOM to train.
training data
The data with which to train the SOM.
length of training
The length of training as measured by the total number of individual data row exposures.
initial learning rate
The initial learing rate for the SOM.
initial neighborhood radius
The initial distance over which a neighborhood is defined.
trained SOM
The ouput trained SOM.
distance metric
The distance metric to use when training the SOM. The Euclidean distance metric uses SOM_PAK as released from the Helsinki University of Technology. The Cosine distance metric uses SOM_PAK as modified by Fareed Qaddoura.
random generator seed
The seed for the random number generator, by default the current time is used.
fixed points
Use fixed point qualifiers in the training.
use weights
Use weighting qualifiers in the training.
read buffer
The number of lines to read at a time.
alpha type
The learning rate function type. Linear (linear) is defined as initial alpha * [1 - (current training step/run length)]. Inverse-time (inverse_t) is defined as initial alpha * [(run length/100)/((run length/100)+current training step)].
snapshot file
A series of files to contain the intermediate SOMs. If the name contains %d the current step number will used in the file name.
shapshot interval
The interval at which to take the snapshots.

Code Reference

uiVsom.vsom(cin, din, cout, rlen, alpha, radius, rand='#', fixed='#', weights='#', buffer='#', alpha_type='#', snapfile='#', snapinterval='#', path='\\bin\\SOM_PAK\\')

Trains a SOM using the specified parameters.

Arguments :
cin

The input codebook file.

din

The training data.

cout

The output codebook file.

rlen

The length of training.

alpha

The initial rate of learning.

radius

The initial radius of neighborhoods.

rand optional

A seed for the random number generator.

fixed optional

Use fixed points.

weights optional

Use weights for variables.

buffer optional

The read buffer size in number of lines.

alpha_type

The type of alpha decrease either linear (default) or inverse time.

snapfile

The snapshot filename.

snapinterval

The interval between snapshots.