Skip to main content
. Author manuscript; available in PMC: 2010 Jul 16.
Published in final edited form as: Neuron. 2005 May 5;46(3):483–492. doi: 10.1016/j.neuron.2005.03.022

Figure 7. Effect of Learning on the Precision of PN Spiking.

Figure 7

(A and B) AL model with facilitation of inhibitory synapses only. (Left) Only the LN input contains noise. (Right) Both PN and LN inputs contain noise. (A) Difference between PN spike phases at consecutive trials with noise (as shown in Figure 6), averaged across all PNs, across ten trial pairs (starting from trial pair 8–9), and finally, across ten independent trial sequences with different noise. Synaptic plasticity reduced the effect of the input noise to LNs by more than 30%. (B) (Top) PN spikes were counted in 10 ms bins during “odor” stimulation. For each bin, a standard deviation of PN spikes across ten trials, <STD>, was calculated starting from trial 10 and then averaged for all PNs in the network. Each trial lasted 500 ms. (Bottom) <STD> was averaged across ten independent trial sequences with different noise. Average <STD> obtained in the model with inhibitory plasticity, <STD>OnlyGABA, was subtracted from the result obtained in the model without plasticity, <STD>NoPlasticity.

(C) Average difference between PN spike phase distributions (left) and average STD (right) for the AL model where all synapses display facilitation during repetitive stimulations. Both PN and LN inputs contain noise. Plasticity in all synapses greatly decreased response variability despite the presence of input noise.