https://github.com/rodrigogiraoserrao/nnni

Neural Networks with No Imports
https://github.com/rodrigogiraoserrao/nnni

framework machine-learning neural-network no-deeplearning-frameworks no-dependencies python python3 vanilla-python

Last synced: about 2 months ago
JSON representation

Neural Networks with No Imports

• Host: GitHub
• URL: https://github.com/rodrigogiraoserrao/nnni
• Owner: rodrigogiraoserrao
• License: mit
• Created: 2021-03-20T19:06:50.000Z (almost 4 years ago)
• Default Branch: main
• Last Pushed: 2021-03-20T23:40:15.000Z (almost 4 years ago)
• Last Synced: 2024-11-03T23:19:21.749Z (3 months ago)
• Topics: framework, machine-learning, neural-network, no-deeplearning-frameworks, no-dependencies, python, python3, vanilla-python
• Language: Python
• Homepage:
• Size: 21.5 KB
• Stars: 0
• Watchers: 1
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# NNNI

### Neural Networks with No Imports (in Python)

As a proof of concept, this repository contains a *single*

Python script with a full implementation of a framework

for artificial neural networks while making use of **no**

imports whatsoever.

This means that the implementation is done purely in

vanilla Python, without even resorting to the standard

library.

Here are some of the things that had to be implemented

by hand because no `import` were allowed:

 - pseudo-random number generation;

 - matrices and matrix algebra:

   - matrix addition and subtraction

   - matrix multiplication and division with scalars

   - matrix comparison with scalars

   - matrix transpose

 - classes for abstract activation and loss functions

 - Leaky ReLU activation function

 - MSE loss function

 - layer and neural network classes

 - forward pass and backpropagation

 > As of now, the code may have a bug that is preventing the

 neural network from learning on the MNIST dataset.

 To be sorted out soon, live on [mathspp.com/twitch](https://mathspp.com/twitch).