Volen Center for Complex Systems, Brandeis University, Dept. of Physics, Harvard University, Dept. of Biology, Brandeis University
Decorrelation of spike trains by synaptic depression
Spike trains recorded in awake, freely behaving monkeys show strong temporal autocorrelations (Baddeley et. al., Proc. Roy Soc. B264:1175, 1997). A more efficient representation of the information in such trains can be constructed by removing these correlations. Using a synapse model based on slice data, we show that synaptic depression removes correlations from realistic spike sequences. This result is verified experimentally by stimulating layer 4 in slices of rat visual cortex and recording field potentials in layer 2/3. Amplitude-weighted correlation histograms of the layer 2/3 responses indicate that correlations in the stimulus sequences are almost completely eliminated. Maximal decorrelation of synaptic transmissions requires the parameters of depression to be matched to the statistics of the presynaptic spike trains. This may be accomplished through presynaptic forms of LTP and modulation that affect short-term plasticity.