Competitive Group

In relation to Competitive Network, a competitive group is an unit of a collection of neurons within a network in which the neurons compete with each other to represent clusters of inputs. This page covers common properties found within a competitive group.

Creation

Logic 

Update Method: Update to current method to Rumelhart-Zipser or Alvarez-Squire.

Epsilon: This field sets the standard learning rate to determine how quickly synapses change.

Winner Value: Insert value for the winning neuron in this field.

Loser Value: Insert the value for all losing neurons in this field.

Use Leaky Learning: By checking this field, all weights learn on each time step, rather than just the winning weight (WTA).

Leaky Epsilon: The learning rate for losing neurons, when leaky learning is enabled.

Normalize Inputs: By checking this field, the inputs would be normalized prior to being used in setting weights.

Synapse Decay Percent: This field sets how much the synapse will decay.

Layout 

Neuon Layout: This field lets the user choose perferred neuron layout.

Layout Style: This field lets the user choose perferred layout style of the neurons. See neuron layout for details.

Spacing: This field sets the distance between the neurons.

Edit

Summary 

Parent: The parent group type, if any.

Incoming

Outgoing

Neurons 

Basic Data:Activation level is set in this field for the group of neurons. Additionally, labelling of the neurons are set in this field.

Update Rule: Neuron type can be updated in this field.

Layout 

Neuron Layout: Select preferred neuron layout in this field.

Horizontal Spacing: The distance between the neurons from left to right

Vertical Spacing: The distance between the neurons up and down.

Manual Columns: Check this field if user wishes to manually set number of columns for the group.

Number of Columns: The number of preferred columns.

Input Data 

User can input data here manually or by uploading .csv file.

Right Click Menu


Edit/Train Competitive Network:  Opens a dialog for editing and training the network.

Rename:  Rename the network in this field.

Remove Network:  Deletes the network and it's constitutes.

Add Current Pattern to Input Data:  By clicking on this selection, it takes the current pattern and input it as the input data.

Train on Current Pattern:  Train the network with the current pattern input data.

Randomize Weights:  Selecting this option randomizes the synaptic weights.