Hebbian layer block

From Piki

The hebbian layer block is essentially a weight layer for use with Hebbian learning.

Hebbian layer
Input ports	1
Output ports	1
Deployable	Yes
Weights	Yes
Memory	No
Interactive GUI	No

Contents 1 Usage 2 Operation 3 Settings 4 GUI details 5 See also

Usage

The hebbian layer is used for associative learning or novelty filtering.

Operation

Settings

The settings can be modified using the settings browser.

Hebbian layer settings

(Layout) Inputs: Number of input features. Outputs: Number of output features. (Settings) Noise Level: Boundary for initial weight randomization. Locked: Determines if the block should ignore reset signals. (Optimization) Brain Damage: Enables or disables adding random noise to weights during training. (If Brain Damage) Damage Noise: Noise as percentage of the weight. (If Brain Damage) Damage Probability: Probability of a weight getting corrupted.

GUI details

In design mode, the block GUI shows the number of inputs and outputs. If there enough display space it will draw all weighted connections:

Hebbian layer design GUI

In training mode the GUI shows the strength of the individual connections. A weak connection with a small weight is transparent while a strong connection is opaque. Positive connections are blue and negative connections are red.

Hebbian layer training GUI

See also

• Block - Article covering general block principles.
• List of Block components - List of all available blocks.