An OOP Deep Neural Network using a similar syntax as Keras with many hyper-parameters, optimizers and activation functions available.
Just clone the repository to your computer to get started:
git clone git@github.com:kinoute/Elyane.git
cd Elyane
The only dependencies to make this neural network work are "numpy" and "pdoc3". They can be installed directly from the main directory if you don't have them already:
pip3 install -r requirements.txt
Three examples are available: mnist.py
, fashion_mnist.py
and xor.py
.
- Batch Gradient Descent
- Mini-batch gradient descent
- SGD
- Full-Connected Layer
- Softmax Layer
- Dropout Layer
- Momentum
- RMSprop
- Adam Optimization
- Amsgrad optimization
- Mean Squared Error (MSE)
- Mean Absolute Error (MAE)
- Cross Entropy
- Multi-Class Cross Entropy
- L1 Regularization
- L2 Regularization
- Dropout
- Sigmoid
- TanH
- Relu
- Leaky Relu
- Softmax
- One hot encoding
- Normalization of images