Help Topics
Concepts
Package
Class
Self-Organizing Map Bean Properties
and Use
The Self-Organizing Map Bean panel provides these options:
- Architecture
- The network architecture consists of these parameters:
- Inputs, which is calculated when beans are generated from the
Training File.
- Output rows, which is the number of rows in the network bean's
output layer.
- Output columns, which is the number of columns in the network bean's
output layer.
- Learn Rate
- Enter a value to control how much the network weights are changed during
a weight update. Larger values cause more change. Learn rate is a real
value between 0.0 and 10.0, with a typical starting value of 0.2.
- Number of Passes
- The value for the number of passes.
Training the network will stop when attained.
- Conscience
- Enter a value to control the bias factor that determines
the distance a losing unit can reach in order to win a
competition. Conscience is a real value between 0.0 and 1.0.
- Agent Mode
- Select one of the following agent modes:
Train implies that the network bean's weights are
unlocked, and network weights will be adjusted
as data is processed.
Test implies that the network bean's weights are locked, and
that error calculations will be performed as data is processed.
Run implies that the network bean's weights are locked and
no error calculations are made.
- Epoch (batch) updates
- Select batch updates if network weights are updated only after a complete
training epoch. Otherwise weights are updated after each record.
The Self Organizing Map Bean panel is used to create a network with specified
architecture and training parameters.
The Mode is set so that the network bean can be trained or used
to provide an independant data source to test that training is sufficient.
Steps in using the panel for training include:
- Set the architecture input value to the number of outputs from
the bean providing data.
- Set the number of output rows and columns.
- Press the Set Architecture button.
- Train the network by pressing the Step, Cycle, or Run buttons on the Agent Editor
toolbar.
- You may wish to press the Stop toolbar button,
change the Maximum RMS Error or Tolerance, and start again. If you change the
network architecture, press Set Architecture for the changes to
take effect. Press the
Reset Weights button to re-initialize the
network weights before starting training again if you wish.