Fig. 6: Synaptic dynamics related to memory formation, consolidation and improvement.
From: Memory consolidation and improvement by synaptic tagging and capture in recurrent neural networks
a During the first seconds after the learning stimulus (“L”), the average early-phase weight (red) of the synapses connecting the stimulated neurons increases, while the average late-phase weight remains constant (blue). A recall stimulus (“R”), provided 10 s after the learning stimulus, does not exert a substantial effect on the synaptic weights. b Several hours after learning, the average early-phase weight decays, while the late-phase increases such that the average total synaptic weight remains on a high level. Providing a recall stimulus 8 h after learning triggers an increase of the average early-phase weight and, thus, of the total synaptic weight. a, b Averaged over ten trials; the error bands (representing the standard deviation) are too small to be visible. c Mean early- and late-phase weights within the assembly and within the non-assembly (control) population immediately before and immediately after the 10s and 8h-recall. A.i. active improvement component (difference between solid blue and blue hatches); P.i. passive improvement component (difference between orange grating and blue hatches). Data from one sample network. Error bars show the standard deviation across the subpopulation. See Supplementary Figs. S4 and S5 for detailed distributions of the underlying data. Parameter setting: wie/h0 = 4, wii/h0 = 4, nCA = 150.
