An accelerator for neural networks with pulse-coded model neurons

G. Frank, G. Hartmann, A. Jahnke, M. Schafer
1999 IEEE Transactions on Neural Networks  
Parallele Rechnernetzwerke in der Produktionstechnik", ME 872/4-1 2. Acknowledgment: We thank Reinhard Eckhorn (Universität Marburg) for many fruitful discussions, theoretical support and participation in his practical experience. We also thank Heinrich Klar (TU Berlin) for cooperative discussions and complementary investigations. Abstract The labelling of features by synchronization of spikes seems to be a very efficient encoding scheme for a visual system. Simulation of a vision system with
more » ... llions of pulse-coded model neurons, however, is almost impossible on the base of available processors including parallel processors and neurocomputers. A "one-to-one" silicon implementation of pulse-coded model neurons suffers from communication problems and low flexibility. On the other hand, acceleration of the simulation algorithm of pulse-coded leaky integrator neurons has proved to be straightforward, flexible and very efficient. Thus we decided to develop an accelerator for a special version of the French and Stein neurons with modulatory inputs which are advantageous for simulation of synchronization mechanisms. Moreover, our accelerator also provides a Hebb'ian-like learning rule and supports adaptivity. Up to 128K neurons with a total number of 16M freely allocatable synapses are simulated within one system. The size of networks, however, is not at all limited by these numbers as the system may be arbitrarily expanded. Simulation speed obviously depends on the number of interconnections and on the average activity within the network. In the case of locally interconnected networks for simulation of vision mechanisms there is only a very low percentage of simultaneously active neurons: stimuli are not simultaneously presented in all orientations and at all positions of the visual field. In these cases our accelerator provides close to real-time behaviour if one second of a biological neuron is simulated by 1000 time slots.
doi:10.1109/72.761709 pmid:18252550 fatcat:nljwq4ckw5gnpln72tz4qvtz7e