Table 1. Parameters used in the integrate-and-fire simulations.
Parameter | Description | Values explored for Figure 5 and 7 | Values explored for Figures 3, 4, 6, 8 |
Learning type | rSTDP, cSTDP | rSTDP, cSTDP for Figs. 3, 4, 6. rSTDP for Fig. 8. | |
A | Depression/Potentiation balance | 0.9, 1.2, 3 | 0.9,1.2 for Figs. 3, 4, 6. 0.9 for Fig. 8. |
D | Synaptic transmission delay | 1, 15 ms | 15 ms |
τSTDP | STDP time constant | 5, 10, 20 ms | 20 ms |
S | Noise level | 1000, 2000 spikes/sec | 2000 spikes/sec |
τsyn | Synaptic time constant | 5, 15 ms | 15 ms |
σinput | Input variance | 100%,200% | 100% |
σnoise | Noise variance | 100%, 200% | 50% |
W min, W max | Minimum/maximum weight | 50 for Figs. 3,4,6; 20 in Fig. 5 | 50 |
F | Target firing rate, for homeostatic scaling | n/a | n/a for Figs. 3–4, 8; 20,80,120 spikes/sec for Fig. 6 |
Relative strength of learning, for homeostatic scaling | n/a | n/a for Figs. 3–4, 8; 0.1,1,10,100 for Fig. 6 | |
ζ | Relative strength of learning, for concurrent plasticity | n/a | n/a for Figs. 3–4, 8; −10,−1,−0.1, 0.1,1,10 for Fig. 6 |
Only parameters that were varied are shown in here; for other parameters that were fixed across simulations, see Methods.