Likewise, a code based on the exact number of spikes elicited by each neuron might be irrelevant in the face of similar fluctuations in response amplitude. In two recent studies,38,39 we have examined the range
of truly possible representation schemes and their feasibility in view of the intense fluctuation in the responses. We have initially shown that in spontaneously developing large-scale random networks of cortical Sorafenib B-Raf neurons in vitro the order in which neurons are recruited following each stimulus is a naturally emerging representation primitive that is invariant to significant temporal changes in spike times. With a relatively small number Inhibitors,research,lifescience,medical of randomly sampled neurons, the information about stimulus position is fully retrievable from the recruitment order. It seems that this is due to the existence Inhibitors,research,lifescience,medical of propagation paths: chains of neuronal stations through which activity is required to pass in order to propagate further into the network, regardless of the status
of membrane and synaptic dynamics, Inhibitors,research,lifescience,medical as can be seen in Figure 4. Figure 4 Demonstration of neuronal stations through which activity is required to pass in order to propagate further into the network. A: Pair-order probability matrices, generated from responses to three different stimulation sources of one network: The matrices … In a second study, we followed the path of a stimulus reconstruction approach to compare systematically the representational efficacy of four types Inhibitors,research,lifescience,medical of popular schemes, two rate-based and two time-based: population-count histogram, spike-count, time-to-first-spike, and rank-order. We found that the nature of response in neural populations
dictates strong correlations Inhibitors,research,lifescience,medical between different response features, which are a priori independent (e.g. rank order of first events and population time histogram are completely orthogonal features of a set of general spike trains38,40), resulting in high redundancy in response features. Thus, all representation schemes perform relatively well under all conditions, with an advantage to either scheme depending on the stimulus properties. Time-based representation schemes are also more stable over long periods of time, under changes induced by the long-term dynamics of the neural assembly. Carfilzomib On the other hand, when classification between temporal features of a given stimulus source is sought, there is an advantage to rate-based representation schemes, which are more sensitive to adaptation processes, and hence contain information with regard to the history of stimulation. We have also found that overlap between groups of receptive sheath neurons (neurons that directly respond to the stimulus and serve as a source for the assembly excitation) is translated to similarity in response pattern and can be thought of as a form of generalization.