The DANA23 (Digital Adaptive Neuro ASIC) neurochip implements a universal neurocontroller. The chip contains a complete feed-forward network with integrated weight storage. Furthermore, there is also a modified backpropagation learning algorithm integrated into the ASIC.
The DANAsim programme is a simulation tool for neural networks developed at the GEMAC mbH and specifically adopted for the DANA23 neuro-ASIC.
With DANAsim, feed-forward networks with up to four layers (one input layer, one or two hidden layers, and one output layer) can be processed. The maximum number of neurones per layer is 64. All neurones have a quadratic activation function.
Learning and testing data for the neural network are to be supplied in the form of an appropriate pattern-file. All learning parameters for a modified form of the standard back-propagation algorithm can be adjusted easily in the program menus.
It is possible to connect the simulator to the DANA23 neuro-ASIC PCI-board, developed at GEMAC mbH. In this way, neural networks developed using the simulator can be tested on corresponding hardware.
