NEURAL ENGINE

A Comprehensive Deep Learning Framework with Automatic Differentiation

Research Software for Neural Network Education & Experimentation

Technologies

Java Script

Used for implementing frontend logic, handling user interactions, and communicating with the Flask backend server for real-time predictions in the web applications.

Python

The entire Neural Engine is implemented in Python, leveraging its rich scientific computing ecosystem that runs seamlessly across Windows, macOS, and Linux.

VS Code

For development, I use Visual Studio Code for its exceptional Python support, including IntelliSense, integrated debugging, and a rich extension ecosystem that streamlines neural network development.

DEMOS

Hand Drawn Digit Recognizer

Showcases drawn digit recognition as well as the inner workings of neural networks.

Neural Networks for Quadratic Equations

Showcases a full-featured web platform with all the tools needed to research and experiment with Neural Networks for Quadratic Equations.
Built from the ground up using Neural Engine.

The full source code and inner workings of the mentioned projects are available on GitHub, accompanied by instructions for running them locally.

GitHub Repo

USAGE

You can start your journey with Neural Engine by following the setup guide in the official Get Started with Neural Engine guide.

ABOUT

The Neural Engine is a first principles deep learning framework with nine activation functions (ReLU, Sigmoid, Tanh, Swish, GELU, and more), robust optimizers (SGD with momentum, Adam), and an end-to-end workflow from preprocessing to training and evaluation. Demonstrated by digit recognition with interactive visualization, a universal alphanumeric recognizer, and a quadratic equation predictor, it delivers numerically stable implementations, automatic differentiation, a modular design, and performance monitoring, bridging theory and practice to make real world pattern recognition and classification accessible.

GitHub

The Developer

Neural Engine is a personal project created, maintained, and documented by a single developer, Matt. It was built with a passion for learning, understanding, and helping others. Through this project, I aim to expand my knowledge, experiment with new ideas, and provide valuable resources to those who might benefit from them.

Got questions?

Neural Engine