ml-cpp

ml-cpp is a machine learning library written in C++ designed for ease of use, flexibility, and performance. The library aims to provide fundamental machine learning algorithms and tools to help users understand and implement machine learning concepts efficiently.

Current Features

• Fully Connected Neural Network (FCNN):

  • Implementation of fully connected layers, allowing users to create deep learning models.

• Activation Functions:

  • Sigmoid activation function for non-linear transformations in neural networks.

• Loss Functions:

  • Mean Squared Error (MSE) loss function for regression tasks.

• Backpropagation:

  • Complete implementation of backpropagation for training neural networks.

• Training Routine:

  • Basic training routine that supports feeding input and target data, adjusting weights based on the computed gradients.

• Prediction:

  • Capability to make predictions after training the model on provided datasets.

Installation

To build and run the ml-cpp project, follow these steps:

1. Clone the repository:

   git clone https://github.com/ismailabouzeidx/ml-cpp.git
   cd ml-cpp

2. Compile the project using CMake

   mkdir build && cd build
   cmake ..
   make -j

3. Run the program

   ./ml-cpp

4. Example Usage

   // Example usage
   NN net(0.01f);
   net.add_layer(std::make_unique<fully_connected_layer>(2, 4));
   net.add_layer(std::make_unique<sigmoid_layer>(4, 4));
   net.add_layer(std::make_unique<fully_connected_layer>(4, 1));
   net.add_layer(std::make_unique<sigmoid_layer>(1, 1));
   
   // Train the network with your input and target data
   net.train(input_data, target_data, epochs);