SAND Kit




This package contains libraries designed for neural networks training & calculations, including linux kernel module for SAND board neuro accelerator.

Calculation library supports both Multi Layer Perceptron (MLP), Radial Basic Function (RBF) & Kohonen networks in hardware acceleration mode and MLP & RBF networks in software emulation mode. It provides C++ class library and simple C routines for neural network calculations. Library have flexible configuration including maximum permited memory usage, calulation & memory usage optimization options, input & output data sources, debugging options. Configuration can be changed simply by editing configuration file and/or with built in gtk+ user interface.

Learning library provides comfort & flexible way to wrote new learning algorithms & methods using powerful modules library, which also included in this package. Learning library also contains several common learning routines including enhanced random search & resilient backpropogation algorithms. It is integrated with calculation library in way to using software as well as hardware modes. All methods have separate configuration files. Support for interface plugins is provide & all learning routines have built in gtk+ interface plugins.

Project contains libraries itself, some documentation and fully functional examples.





SourceForge Logo