nano-keras

Overview

nano-keras is a deep learning library written in Python using NumPy. It's designed to handle the creation and training process of most neural network types, allowing you for quick and easy prototyping and deployment.

The project is heavily inspired by Keras, the most popular deep learning API in the world, as I'm trying to implement my library in simmilar style and functionality to Keras

Key Features

- Simplicity: Built using Python and NumPy, making it easy to read and understand each part

- Educational: Intended as a learning tool to understand neural network components at a lower level

- Customization: Allows for tinkering and understanding the core mechanics of neural network operations

What you can find in nano-keras

Layers: Dense, Dropout, Reshaping layers, Convolutional layers, Pooling layers and Recurrental Layers

Optimizers: SGD, Adam, Adadelta, Adagrad, RMSProp, NAdam and much more

Activation functions: Sigmoid, Tanh, ReLU, ELU, LeakyReLU, Softmax

Loss functions: MAE, MSE, BCE, CCE, Hinge, Huber

Callbacks: EarlyStopping, LearningRateScheduler, CSVLogger

And much more, you can find all the implemented items in here

Instalation

nano-keras is available on PyPI so in order to download it open a terminal and paste:

pip install nano-keras

You now should have succesfully installed nano-keras so to use it in your python file you only need to import it like this:

import nano_keras

If you have an issue message me on github or send me an email

Documentation

Documentation is under development and should be finished in the next few days

You can access it here

License

This project is licensed under the MIT License - see the LICENSE file for details

Special thanks

I'd like to thank my teacher, Mateusz Kozlowski, who inspired me to start working on this project and kept me motivated to finish this and evryone who showed support for me

Without you this project would've never come to life