Preprint
Jul 2026
Building a Neural Network from Scratch: Implementation, Evaluation, and Optimization
This paper presents a self-contained neural network framework implemented entirely from scratch -- without relying on automatic differentiation or pre-built deep learning modules, including multi-layer architectures, diverse activation functions, regularization techniques, and state-of-the-art optimizers.
Yuanzhe Jia
· 0 citations