Figure 5.

Results of spike timing-dependent plasticity (STDP) applied to (A) ORN-LN and (B) LN-LN connections in the MGC AL neuronal model. Adapted weights of both connection types are shown to converge from an initial randomized state to stable asymptotic values over many epochs. (C) Normalized synaptic weights of ORN-LN connections show that LN input tunings (blue circles) covers the space of possible input ratios (shown as linear segments) after STDP learning. (D) PN firing rate trajectories to three different ORN1:ORN2 input ratios (1:3, 1:1, and 3:1) both before and after STDP adaptation show greater attractor separation after learning (3 dimensional PCA explains 90% of total variance).