Description

I'd like to introduce you to Neural, a small project I developed in a week. It's a C++ tool for experimenting with artificial neural networks. It's designed to be easy to use and straightforward to set up (simply create the Docker container and start coding!). If you encounter any issues, please don't hesitate to let me know. The tool works with a Docker image, allowing for quick installation and configuration of CUDA and OpenGL, as well as the necessary tools for data visualization. To access the installation guide, API documentation, and tools, please visit our website.

Website.

Exemple

MNIST

Use of the stochastic gradient descent method with the mean square error calculation function.

Roadmap

• Activation function
  • Tanh
  • Sigmoid
  • Relu
  • LeakyRelu
  • Softmax (in dev)
• Loss function
  • MSE
  • CRE
• Layer
  • Full connected
  • Flatten
  • Convolution
  • Associative (in R&D)
• CudaSupport
  • Matrix multiplication
  • Cross correlation
  • Convolution
• Save&Load networks
  • JSON
  • HDF5
• MathUtils
  • Cross correlation
  • Padding
  • Convolution
• GUI (In dev)

License

Distributed under the MIT License. See LICENSE for more information.

Contact

Nicolas Brugie - [email protected]