muchlakshay/Dual-Backend-MLP-From-Scratch-CUDA

A fully from-scratch Multi-Layer Perceptron built in CUDA C++ with support for both GPU and CPU training. Includes multiple activation and loss functions, a clean and modular architecture, and an easy-to-use API, all without relying on external machine learning libraries.

1 best points · 2 comments · 1 Hacker News post · 21 GitHub stars · View on GitHub

Hacker News discussions

• Show HN: I Am 15 and Built a Dual Backend MLP from Scratch Using CUDA C++ (1 points, 2 comments, by muchlakshay, 2025-07-23)