An intensive, feature-rich deep learning library written completely in C++, along with python interface for user-friendly accessibility.
The following features are currently supported:
- Optimizers: Features variety of optimizers like SGD, SGD with Momentum, SGD with Nesterov Momentum, Adagrad, Adam, Nadam, Adamax, AMSGrad, Adadelta, RMSProp, Adabound, AMSBound
- Activation functions: Includes several activation functions such as Sigmoid, ReLU, Leaky ReLU, ELU, Tanh, Softmax
- Loss functions: Supports various loss functions like Mean Absolute Error, Mean Squared Error, Quadratic, Half Quadratic, Cross Entropy, NLL
- Weight initializers: Provides multiple weight initializers like Random, Xavier Uniform, Xavier Normal, LeCun Uniform, LeCun Normal, He Uniform, He Normal
- Regularizers: Contains different regularizers such as L1, L2, L1L2, None
- Layers: Fully Connected(Dense) Layer
The examples provided here give a general idea about creating, training, evaluating and performing operations like save/load on a model using both C++ and Python.
Detailed structure follows below:
- Creating a model
- Training a model
- Evaluating a model
- Saving and loading a model
Detailed structure follows below:
- Creating a model
- Compiling a model
- Training a model
- Evaluating a model
- Saving and loading a model
This program is licensed under GNU General Public License version 3.0 (GPLv3) and is free for redistribution and/or modification as per the terms of the license. The license file is available in the repository.