Browse all publications by topic
Browse all publications by year
- A. Hilmann, S. Heller,
U. Roth, C. Piepenbrock, K. Obermayer, and K. Klar. A Graphical
Programming and Simulation Environment for Integrate-and-Fire Neuronal
Networks.
.
In European J. Neurosci., page 404. European Forum of Neuroscience
ENA, 1998.
We have developed a simulator for networks of integrate-and-fire
neurons as part of a larger project to implement such networks directly in
hardware. The model neurons are arranged in layers. They have independent
variable thresholds for activity and learning. Synaptic connections between
the neurons may be defined individually or by arbitrary pattern functions.
Each connection can be configured with respect to its propagation time delay,
its PSP time behavior, and its adaptability by different Hebbian learning
rules. The software package uses a modular description language for the
network topology, the neuron parameters, the learning rules, as well as for
the input stimulation protocols. It provides (i) a compiler for the modular
network description files that generates network connectivity and parameter
files, (ii) a simulator to run network simulations, and (iii) an interactive
graphical user interface to control the simulator and visualize the network
neurons with their membrane potentials, activities, learning behavior, and
their synaptic connection weights. Furthermore, it implements statistical
functions for the analysis of the network and its time behavior. The
simulator may be run in batch mode without the user interface--and it could
later be replaced by the microchip designed to implement exactly the same
functionality in hardware. The package is implemented in ANSI C and JAVA and
is designed for maximum portability across platforms. We demonstrate its
capabilities in simulations of figure-ground separation and path
finding.
|
