Explicit list of all implemented features

Neural Network Models and Architectures

• Layers
  • Fully Connected layer
  • Constitutional layer
  • Recurrent layer
• Activation Functions
  • Sigmoid
  • Improved Sigmoid
  • Tanh
  • ReLU
  • Gaussian
  • Identity

Learning Algorithms and Optimizations

• Back-propagation algorithms
  • Stochastic gradient descent with Momentum
  • Adam
• Layer optimizers
• Dropout

Data

• Multiple data type managed
  • Simple data
  • 2D data
  • Temporal data
  • Time Series
• automatic normalization of data

Other Features

• Windows and Linux compilation
• Save and Load neural network from file

Tests

• Unit tests
  • Linux with GCC
  • Windows with MSVC
• Datasets tested
  • Iris
  • Wine
  • MNIST
  • Fashion-MNIST
  • CIFAR-10