Subsystems

You can create the model of your Survey ART neural network on the Subsystems tab. This gives you the list of subsystems in a table form, where you can create, edit and delete subsystems.

Use the Add button to create new subsystems.

Existing subsystems can be edited using the Edit button.

To delete a subsystem, select one or more subsystems using the mouse, then click on the Delete button.

Adding and Editing Subsystems

Once you have clicked either on the Add or the Edit button, you will access the Subsystem Editor. You can set up subsystem parameters here.

Name Enter a name for the subsystem. Maximum 50 characters are allowed.

Type Select the subsystem type. You can choose from the following subsystem types:

General
This is the most common type. General subsystems can establish associations with other general subsystems.

Evaluation
This subsystem evaluates and combines the signal coming from the general subsystems.

Choice
The choice subsystem will provide the final decision.

Max. Features The maximum number of features for the dictionary entries in the subsystem. Actually this is the dimension of the activation vectors in the ART neural network. If you change the number of features in an existing subsystem, the Resize dialog will appear.

Max. Categories The maximum number of categories that can be created in the subsystem during the learning process.

Dictionary This is a list of dictionary entries for the current subsystem, which is only visible if you edit an existing subsystem. You can use the Add, Edit and Delete buttons to manage the dictionary entries.

Resizing Subsystems

If you have modified the number of features in an existing subsystem, then you have to choose how to resize the subsystem. If you have reduced the number of features you have to select some of the vector elements to remove. If you have increased the number of features, then have to select where to insert the new vector elements.

Editing Clauses

Enter the details for the dictionary entry. Define the activation vector and an appropriate textual representation for it.

Clause Enter a textual representation for the actual activation vector. Maximum 255 characters are allowed.

Code This is the activation vector for the textual clause given above. Enter N real values in the interval [0..1], separated with commas, where N is the number of features in the subsystem. You can optionally enclose the vector in square brackets. Examples: 0,0,0,1 or [1,0.5,0,0.9]

Messages

You can define message groups and messages here. Messages groups are merely organizational units, they help to work with messages. The subjects will be trained using these messages.

Use the Add button to add a new message group.

Existing message groups can be edited using the Edit button.

To delete a message group, select one or more groups using the mouse, then click on the Delete button.

Adding and Editing Message Groups

Once you have clicked either on the Add or the Edit button, you will access the Message Group Editor.

Name Enter a name for the message group. Maximum 50 characters are allowed.

Messages This is a list of messages for the current message group. This table is only visible if you are in edit mode. You can use the Add, Edit and Delete buttons to manage the messages. You can change the position of a message using the Move Up and Move Down buttons.

Adding and Editing Messages

This panel is used to set up a message. You can define the following parameters:

Type The message type defines how the subject learns the message during the training phase. You can choose from two messages types:

Simple
If the message is a simple message, then both parts of the message are learnt equally.

Causality
If the message expresses some kind of causality, then the first part of the message is considered more important.

Source This option defines how reliable is the source of the message. This modifies the learning rate during training. The following values are available:

Distrusted
Learning rate will slightly decrease.

Neutral
No change will occur in the learning rate.

Trusted
Learning rate will slightly increase.

Left Clause This list contains the dictionary entries from all subsystems, from which you can select the clause appropriate for the message.

Right Clause This list contains the dictionary entries from all subsystems, from which you can select the clause appropriate for the message.

Subjects

This tab gives you a list of subjects in the experiment. You can choose from the following tasks:

Use the Add button to create new subjects.

Existing subjects can be edited using the Edit button. You can also modify the ART and ARAM parameters for each subject.

To delete a subject, select one or more subjects using the checkboxes, then click on the Delete button.

Use the Training button to select the list of message groups to train to the selected subjects.

Click here to analyse the neural network for each subject during or after training.

This button brings up the Question window, where you are allowed to ask questions from the selected subjects.

Adding and Editing Subjects

Once you have clicked either on the Add or the Edit button, you will access the Subject Editor. You can set up the subject parameters here.

No. of subjects This field is only enabled if you add a new subject. Enter the number of subjects you want to be generated. If you add more than one subject, then an index will be appended to the end of each subject's name.

Name Enter a name for the subject. Maximum 50 characters are allowed.

Vigilance This is the default vigilance value for the subject. Enter a real value in the range [0..1].

Learning Rate This is the default learning rate for the subject. Enter a real value in the range [0..1].

Choice Param. This is the default choice parameter for the subject. Enter a real value in the range (0..1].

Contribution This defines the default contribution for the subject between subsystems. Enter a real value in the range [0..1].

ART params This tab is only visible if you edit an existing subject. You can set up the ART parameters for each subsystem here.

ARAM params This tab is only visible if you edit an existing subject. You can set up the ARAM parameters for each subsystem here.

Editing ART Parameters

This dialog is used to define the parameters for the ART neural network.

Vigilance This is the vigilance value for the subsystem. Enter a real value in the range [0..1].

Learning Rate This is the learning rate for the subsystem. Enter a real value in the range [0..1].

Choice Param. This is the choice parameter for the subsystem. Enter a real value in the range (0..1].

Editing ARAM Parameters

This dialog is used to define the parameters for the ARAM neural network.

Vigilance This is the vigilance value for the subsystem. Enter a real value in the range [0..1].

Learning Rate This is the learning rate for the subsystem. Enter a real value in the range [0..1].

Choice Param. This is the choice parameter for the subsystem. Enter a real value in the range (0..1].

Contribution This defines the contribution for the subsystem pair. Enter a real value in the range [0..1].

Training Window

The selected subjects will learn the message groups that you select in this window.

Overwrite previous data If this option is checked, the subject's neural network data will be cleared before training.

Message Groups The list of message groups to learn. You can use the Add, Edit and Remove buttons to manage the message groups. To change the position of a message group use the Move Up, Move Down buttons.

Selecting Training Groups

This window is used for message group selection for the current training.

Message group Select a message group from the list to train.

Perfect learning If perfect learning is enabled, then all messages in this groups are learned perfectly, and type and source parameters are not taken into account.

Training Progress

This panel allows you to control the steps of the training process. It displays the messages in the training queue.

Use the Learn Message button to train a single message.

Use the Learn Message Group to train all messages from a message group.

Use the Learn All button to train all messages in the queue.

The Stop button can be used to stop the current training. It is only enabled if training is in progress.

Connections

This screen shows the connections between the subsystems. A thicker line means there are more established connections between the two subsystems.

Examine Window

This window can be used the view the neural network data. Select the ART or ARAM tabs, then click on a row in the table to view the values. Use the Show Connections button to view the number of established categories between subsystems graphically.

ART Data Window

This screen shows the prototype values in the ART neural network. Each green block represents a vector element. A taller block means a higher value.

ARAM Data Window

This screen shows the prototype values in the ARAM neural network. Each green block represents a vector element. A taller block means a higher value.

Question Window

You can ask questions from the selected subjects in this window.

Question Select a message clause from the list. Subjects will use this clause to search for associations with other subsystems.

ART Vigilance The vigilance of the ART subsystem, which creates the category prototype from the question.

Direct Choice Left Vig. The left vigilance parameter for the ARAM module when searching for a direct link between the question subsystem and the choice subsystem.

Direct Choice Right Vig. The right vigilance parameter for the ARAM module when searching for a direct link between the question subsystem and the choice subsystem.

ARAM Left Vigilance The left vigilance parameter for the ARAM module when searching for associations between general subsystems.

ARAM Right Vigilance The right vigilance parameter for the ARAM module when searching for associations between general subsystems.

Evaluation Subsystem This list contains the evaluation subsystems, whose task is to evaluate the signal coming from the general subsystems. Select one to use in this session.

ARAM Left Vigilance The left vigilance parameter for the ARAM module when searching for an association in the evaluation subsystem.

ARAM Right Vigilance The right vigilance parameter for the ARAM module when searching for an association in the evaluation subsystem.

Choice Subsystem This list contains the choice subsystems, whose task is to provide a final decision. Select one to use in this session.

ARAM Left Vigilance The left vigilance parameter for the ARAM module when the combined signal is activating the choice subsystem.

ARAM Right Vigilance The right vigilance parameter for the ARAM module when the combined signal is activating the choice subsystem.

Results Window

This window shows the results of the question. Each row shows a single phase, and the answer is in the last row.

	Use the Add button to create new subsystems.
	Existing subsystems can be edited using the Edit button.
	To delete a subsystem, select one or more subsystems using the mouse, then click on the Delete button.

Name	Enter a name for the subsystem. Maximum 50 characters are allowed.
Type	Select the subsystem type. You can choose from the following subsystem types: General This is the most common type. General subsystems can establish associations with other general subsystems. Evaluation This subsystem evaluates and combines the signal coming from the general subsystems. Choice The choice subsystem will provide the final decision.
Max. Features	The maximum number of features for the dictionary entries in the subsystem. Actually this is the dimension of the activation vectors in the ART neural network. If you change the number of features in an existing subsystem, the Resize dialog will appear.
Max. Categories	The maximum number of categories that can be created in the subsystem during the learning process.
Dictionary	This is a list of dictionary entries for the current subsystem, which is only visible if you edit an existing subsystem. You can use the Add, Edit and Delete buttons to manage the dictionary entries.

Clause	Enter a textual representation for the actual activation vector. Maximum 255 characters are allowed.
Code	This is the activation vector for the textual clause given above. Enter N real values in the interval [0..1], separated with commas, where N is the number of features in the subsystem. You can optionally enclose the vector in square brackets. Examples: 0,0,0,1 or [1,0.5,0,0.9]

	Use the Add button to add a new message group.
	Existing message groups can be edited using the Edit button.
	To delete a message group, select one or more groups using the mouse, then click on the Delete button.

Name	Enter a name for the message group. Maximum 50 characters are allowed.
Messages	This is a list of messages for the current message group. This table is only visible if you are in edit mode. You can use the Add, Edit and Delete buttons to manage the messages. You can change the position of a message using the Move Up and Move Down buttons.

Type	The message type defines how the subject learns the message during the training phase. You can choose from two messages types: Simple If the message is a simple message, then both parts of the message are learnt equally. Causality If the message expresses some kind of causality, then the first part of the message is considered more important.
Source	This option defines how reliable is the source of the message. This modifies the learning rate during training. The following values are available: Distrusted Learning rate will slightly decrease. Neutral No change will occur in the learning rate. Trusted Learning rate will slightly increase.
Left Clause	This list contains the dictionary entries from all subsystems, from which you can select the clause appropriate for the message.
Right Clause	This list contains the dictionary entries from all subsystems, from which you can select the clause appropriate for the message.

	Use the Add button to create new subjects.
	Existing subjects can be edited using the Edit button. You can also modify the ART and ARAM parameters for each subject.
	To delete a subject, select one or more subjects using the checkboxes, then click on the Delete button.
	Use the Training button to select the list of message groups to train to the selected subjects.
	Click here to analyse the neural network for each subject during or after training.
	This button brings up the Question window, where you are allowed to ask questions from the selected subjects.

No. of subjects	This field is only enabled if you add a new subject. Enter the number of subjects you want to be generated. If you add more than one subject, then an index will be appended to the end of each subject's name.
Name	Enter a name for the subject. Maximum 50 characters are allowed.
Vigilance	This is the default vigilance value for the subject. Enter a real value in the range [0..1].
Learning Rate	This is the default learning rate for the subject. Enter a real value in the range [0..1].
Choice Param.	This is the default choice parameter for the subject. Enter a real value in the range (0..1].
Contribution	This defines the default contribution for the subject between subsystems. Enter a real value in the range [0..1].
ART params	This tab is only visible if you edit an existing subject. You can set up the ART parameters for each subsystem here.
ARAM params	This tab is only visible if you edit an existing subject. You can set up the ARAM parameters for each subsystem here.

Vigilance	This is the vigilance value for the subsystem. Enter a real value in the range [0..1].
Learning Rate	This is the learning rate for the subsystem. Enter a real value in the range [0..1].
Choice Param.	This is the choice parameter for the subsystem. Enter a real value in the range (0..1].

Overwrite previous data	If this option is checked, the subject's neural network data will be cleared before training.
Message Groups	The list of message groups to learn. You can use the Add, Edit and Remove buttons to manage the message groups. To change the position of a message group use the Move Up, Move Down buttons.

Message group	Select a message group from the list to train.
Perfect learning	If perfect learning is enabled, then all messages in this groups are learned perfectly, and type and source parameters are not taken into account.

	Use the Learn Message button to train a single message.
	Use the Learn Message Group to train all messages from a message group.
	Use the Learn All button to train all messages in the queue.
	The Stop button can be used to stop the current training. It is only enabled if training is in progress.

Question	Select a message clause from the list. Subjects will use this clause to search for associations with other subsystems.
ART Vigilance	The vigilance of the ART subsystem, which creates the category prototype from the question.
Direct Choice Left Vig.	The left vigilance parameter for the ARAM module when searching for a direct link between the question subsystem and the choice subsystem.
Direct Choice Right Vig.	The right vigilance parameter for the ARAM module when searching for a direct link between the question subsystem and the choice subsystem.
ARAM Left Vigilance	The left vigilance parameter for the ARAM module when searching for associations between general subsystems.
ARAM Right Vigilance	The right vigilance parameter for the ARAM module when searching for associations between general subsystems.
Evaluation Subsystem	This list contains the evaluation subsystems, whose task is to evaluate the signal coming from the general subsystems. Select one to use in this session.
ARAM Left Vigilance	The left vigilance parameter for the ARAM module when searching for an association in the evaluation subsystem.
ARAM Right Vigilance	The right vigilance parameter for the ARAM module when searching for an association in the evaluation subsystem.
Choice Subsystem	This list contains the choice subsystems, whose task is to provide a final decision. Select one to use in this session.
ARAM Left Vigilance	The left vigilance parameter for the ARAM module when the combined signal is activating the choice subsystem.
ARAM Right Vigilance	The right vigilance parameter for the ARAM module when the combined signal is activating the choice subsystem.