C++ Neural Network with Training Visualizer

A C++ implementation of a Neural Network, along with a real-time training visualizer and data generation tool. The project aims to visualize the error, gradient, and weight values as the neural network trains over iterations.

Features

• A simple feedforward ANN with training using backpropagation and activation functions.
• Real-Time Training Visualizer that visualizes error, gradients, and weights over iterations using a Python script.
• Random Training Data Generator: A Python script to generate random training data in the required format.
• Dual File/Terminal Output Logging.

How It Works:

• A Comprehensive and in depth analysis of my Neural Net implementation is available in a PDF format written in LaTeX.
• To use the Training visualizer simply generate training data using the Training data generation tool, manually build the Neural Net, run the Training visualizer and run the Neural Net, this will initiate the training process using the newly generated data set and the training visualizer will update every 100ms as the Neural Net runs through the entire data set.
• The Training data generation tool allows you to generate N amount of data lines, allows you to edit the topology of the Neural Net as well as the range of the data values.

Requirements

• C++17 and newer: For building and running the neural network.
• Python 3.x: Required for the visualizer and data generation scripts.
• Matplotlib & Numpy.

Installation

Note:

• Build instructions can be found below including installing python, matplotlib and numpy using the Python installer, Homebrew and apt for Windows, MacOS and linux respectively. Other build instructions and compiler flags are included as well.

Installing Python

Windows

1. Download the latest Python installer from python.org
2. Run the installer and check "Add Python to PATH" during setup.
3. Verify installation by running in Command Prompt:

   python --version

macOS

1. Open Terminal and install Python using Homebrew:

   brew install python

2. Verify installation:

   python3 --version

Linux (Debian/Ubuntu)

1. Open Terminal and run:

   sudo apt update
   sudo apt install python3 python3-pip

2. Verify installation:

   python3 --version

---

Installing Required Python Libraries

Once Python is installed, install the required libraries (matplotlib and numpy):

pip install matplotlib numpy

Clone the Repository and Build

1. Clone the repository:

   git clone https://github.com/000x999/Neural_Net
   cd Neural_Net

2. Compile and run the program:

   Make sure to have 'Make' installed, 'GCC'/'G++' 
   Set your STL C++ header files 'Include' folder path in the CXXFLAGS section inside the Makefile.
   Simply run 'make' directly in the command line from within the 'Neural_Net' folder.
   To run any of the two python scripts, be sure to have numpy and matplotlib installed,
   then simply run 'py [scriptName.py]' directly in the command line