lightnet

lightnet is a simple implementation of convolutional neural network for myself to understand the basic ideas about how to build a neural network, especially about back propagation algorithm and optimization.

Features

• The core of this project is less than 1000 lines, all in one header file.
• Only use c++ standard library as dependency.
• OOP style and easy to read.
• API copies PyTorch a lot, so you may feel familiar : )
• Simple examples for classification and regression, also has MNIST classification.

Reference

• simple_cnn Idea and codes for MNIST dataset preparation.
• darknet