Skip to content

BinarySemaphore/NeuralNetPy

Repository files navigation

NeuralNetPy

Neural Net from scratch in Python

Modules

Generate

Genetic Algorithm implementation, for those brute force situations

Neural

Neural Network implementations, including trainer (MB-GD / SGD)

Trainer Modules

bp indicates trainer using backpropagation

evo indicates using genetic algorithm

Training Sets

MNIST

Collection of handwritten digits from MNIST

Modules:

  • mnist_data: loads MNIST from ubyte files and can draw in terminals
  • bmp: a 16-bit BMP file reading module

Includes some user recorded digits BMP files

Sentences

Generates random word sentences.

Includes a JSON file of pre-generated sentences with labels 0 or 1 if it includes test (1) or not (0)

Saved Networks

Each saved network directory has trainers to indicate where a trained networks accuracy was achived, so it can be reproduced.

DR Net (dr_net)

Digit Recognition Network (trained on MNIST)

About

Neural Net from scratch in Python

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages