A complete neural network built entirely in x86 assembly language that learns to recognize handwritten digits from the MNIST dataset. No frameworks, no high-level languages - just pure assembly - ~5.3× faster than NumPy

🔬 The most atomic GPT-2 implementation in 265 lines of pure Python & CUDA. A bilingual "Rosetta Stone" for understanding LLM internals from scratch. No dependencies, just math and kernels.

Deep Learning Framework only using numpy: Linear, Convolution, Flatten, Max and Mean Pooling layers, activation functions, Softmax, MSE and Cross Entropy. Useful functions: train, save an load a model to deploy it, get nb of parameters, draw learning curves

📖 ⟪밑바닥부터 시작하는 딥러닝2⟫ 스터디 기록 저장소

A lightweight, deep learning library written in pure Python

A coding exercises with numpy and pytest for Deep Learning From Scratch

A Python project implementing a neural network framework from scratch using NumPy. Includes fully connected (Dense) layers, ReLU and Sigmoid activations, a simple SGD optimizer, and a minimal training loop. Designed for hands-on learning of neural network fundamentals without relying on any deep learning frameworks

Written in pure numpy. "Implementing Transformers with NumPy matrix operations, this book dissects the theory into code." An essential learning roadmap for developers and researchers who want to open the black box and fully understand the principles of deep learning.

밑바닥부터 시작하는 딥러닝 스터디 요약 저장소 (HW)

Neural Network algorithms, concepts and application developed from scratch in python using just numpy, scipy and matplotlib libraries.

A from scratch PyTorch implementation of DDPM and DDIM samplers trained on 2D manifold data. Features a custom linear noise scheduler, sinusoidal time embeddings, and zero external diffusion libraries.

MNIST Handwritten Digits Classifier From Scratch In C++

A from-scratch implementation of feedforward neural networks using NumPy. Developed for the Artificial Intelligence Fundamentals course at the University of Parma, featuring manual backpropagation, mini-batch SGD, and inverted dropout on the MNIST dataset.

A fully vectorized Deep Neural Network (DNN) implementation built from scratch using only NumPy - no deep learning frameworks involved. Covers forward/backward propagation, activation functions, modular architecture, and training with different optimizers - a hands-on deep dive into the fundamentals of deep learning.