Feedback Query Manager

Stores queries of feedback(float* fb_input) necessary for training the neural network in question.

FeedbackQuery

A Feedback Query, is composed of the following:

• neuron(Pointer to FloatUnitNeuron)
  Stores the reference/memory address to a neuron
• fb_input(Array of 2 floating point values)
  Stores values that becomes the input of the feedback method

Structure and Functionalities

A Feedback Query Manager instance includes the following:

Public Methods

• FeedbackQueryManager()
  Creates an instance of the class
• add_query(FeedbackQuery _new)
  Adds a FeedbackQuery instance to an internal array in the current class
• void execute_all()
  Recursively executes all queries stored in the internal array in the current class
• print_current_queries()
  Prints all queries stored in the internal array in the current class

Private Variables

(Skip this section if you just want to use this preexisting class without custom modifications)

• query_list(Private array of FeedbackQuery instances)
  Stores queries of feedback(float* fb_input) generated by execute() of a FloatGlobalOperator instance and feedback(float* fb_input) executed by its corresponding FloatUnitNeuron instance.
• num_query(Int variable)
  Number of queries stored in query_list
• fb_futures (std::vector of std::future)
  Vector of feedback queries to be executed asynchronously
  CURRENTLY NOT IN USE.

Code

The class is defined in the feedback_query/fb_query_manager.hpp and feedback_query/fb_query_manager.cppfile.

Additionally, an option for executing queries asynchronously via multithreading is available. If you wish to execute using multithreading, change the value of ASYNC macro defined in the cpp file from 0 to 1.