Back to my homepage.
Abstract:Networks
of spiking neurons can realize pattern recognition tasks by encoding the
patterns in the synaptic strength of connections between neurons. We examine a
system that receives as an input a linear combination of its stored patterns and
outputs the coefficients of the linear combination. Both the input and the
output are given in temporal coding, hence the computation is based on the
explicit firing times of the input neurons and not on the more common firing
rate of neurons.
The precise formulation imposes several constraints on the system, which
can be only approximately fulfilled by biological neurons. By simulations we
show that the above described approach can be nevertheless realized by
biological neurons in a very reliable way, giving rise to a very simple and
straightforward method for pattern recognition in biological neural systems.
Back to my homepage.