Figure 1
From: Requirements for the robust operant conditioning of neural firing rates
Effective STDP learning windows that show the synaptic change, ΔK , caused by a spike pair with the timing difference, Δt , between the pre- and post-synaptic spikes. The different windows are for different reward signal levels (y = 0, 1, 2, 3, 4, 5, and 6, shown in green, blue, purple, magenta, red, orange, and yellow lines, respectively). A: Classical RSTDP has no LTD or LTP when there is no reward and increasing amounts of both as reward increases. B: Generalized RSTDP allows separate modulation of LTP and LTD and non-zero LTP and LTD when there is no reward. Shown here is generalized RSTDP parameterized to match the effect dopamine has been observed to have on STDP [4]. Switching of LTD to LTP for high reward values is exhibited.