Float Unit Neurons

Individual linkable class instances that represent the individual neurons in a neural network.

Structure and Functionalities

An abstract unit neuron class includes the following:

• state (public floating point value)
  Indicates the current output signal of the neuron
• previous (public array of unit neurons' pointers)
  Stores memory addresses of neurons that the current neuron receives external signals from
• void feedforward() (public virtual method)
  Executes feedforward operation
• void feedback(float* fb_input) (public virtual method)
  Executes feedback operation using fb_input
  Each weights (stored in memory) are updated asynchronously
  Optionally generates a query for the next neuron to operate the feedback operation
  Generated query will be stored in an external object instance of FeedbackQueryManager
• memory (protected floating point array)
  Stores values used for calculating forward and feedback loops
  Example: postsynaptic weights

Code

The abstract class is defined in the unit_neuron.hpp file.