Binary Neuron

This is a common type of activation function in neural networks. The activation level a of a binary node is set to an upper value u (set in the upper bound field) if the weighted input W plus a bias term b is greater than the threshold θ, and a lower value l (set in the lower bound field) otherwise:

This type of activation function is also referred to as a threshold function, and these units are sometimes called "linear threshold units."

Threshold

The value, denoted by θ above. This provides a simple model of the voltage potential that real neuron membranes must exceed in order to fire an action potential.

On Value

The value that turns on neuron.

Off Value

The value that turns off neuron.

Bias

A fixed amount of input to this node, denoted by b above.