SandLib is a collection of libraries designed for neural networks training and use, including a linux kernel module for the SAND neuro acceleration board. The package also contains fully functional sample programs and GTK+ configuration plugins. The calculation library supports multi-layer perceptron (MLP), radial basis function(RBF), and kohonen networks in hardware acceleration mode. It also supports MLP and RBF networks in software mode. The learning library contains random search and resilient back-propagation algorithms. Support is provided for configuration plugins.
