Network: Comput. Neural Syst. 8 (1997) 355–371.A theoretical model for analogue computation in networks of spiking neurons with temporal coding is introduced and tested through simulations in GENESIS. It turns out that the use of multiple synapses yields very noise robust mechanisms for analogue computations via the timing of single spikes in networks of detailed compartmental neuron models.
In this way, one arrives at a method for emulating arbitrary Hopfield nets with spiking neurons in temporal coding, yielding new models for associative recall of spatio-temporal firing patterns. We also show that it suffices to store these patterns in the efficacies of excitatory synapses.
A corresponding layered architecture yields a refinement of the synfire-chain model that can assume a fairly large set of different stable firing patterns for different inputs.Recent experimental results from neurophysiology have shown that in many biological neural systems not only the firing rate, but also the spatio-temporal pattern of neuronal firing carries important information.
Other recent experimental results indicate that it is in fact questionable whether biological neural systems are able to carry out analogue computation with analogue variables represented as firing rates. Due to ‘synaptic depression’ the amplitude of postsynaptic potentials arising from a presynaptic neuron u tends to scale as 1=f where f is the firing rate of u (see, e.g. , Abbott et al 1997). Therefore both slowly firing neurons and rapidly firing neurons u inject roughly the same amount of current into a postsynaptic neuron during a given time window. This suggests that both a McCulloch–Pitts neuron and a sigmoidal neuron model overestimate the computational capability of a biological neuron for rate coding.
In addition, it has been argued that in view of the rather low firing rates of cortical neurons analogue computations in multi-layer neural systems with intermediate variables represented as firing rates would be much too slow to achieve the experimentally observed computation speed of concrete cortical.
