Neural Network Framework is a C++ library and QT 4.x tools that allows for the development of diverse neural networks.
At the lowest level, users can build complex architectures with no restrictions. This level is meant for expert users who can contribute to NNFW by introducing new functionalities. The intermediate level offers classes for quick creation of common neural networks such as multilayer feed-forward network, simple recurrent networks (e.g., Elman, Jordan), and Radial-basis network, among others. The possibility of representing a neural network with XML files provides a general and abstract interface for accessing all the functionalities of NNFW.
The graphical interface is the most user-friendly and easy way to use the framework. It allows users to experiment with complex neural networks, reducing the learning curve.
To use NNFW, your system must meet specific requirements, including CMake, Doxygen, and QT 4.x Library.
If you're using Linux, you can download the source package and unpack it into the preferred directory. There are three types of procedures to compile NNFW: Quick, Normal, and Recommended procedure. The recommended procedure is the most convenient and keeps the main source tree clean.
In this release, a C binding has been implemented, among other enhancements. Overall, NNFW is a reliable software for developing neural networks with a range of features that suit various user levels.
Version 1.1.5: N/A