Figure 7. Decision-making performance.

In all but one network, learning is required for reliable decision-making. A. One out of 25 initial networks generate decisions above criterion. B&C. Triplet but not Standard STDP generates one network near criterion for reliable decisions but no networks demonstrate reliable decisions. D. LTPi alone demonstrates the strongest decision-making networks performances, consistent with its strong stimulus-pair selectivity. E&F. The addition of LTPi to standard or triplet STDP results in networks capable of reliable decisions and more near threshold. These results demonstrate that LTPi is required necessary for reliable decisions in addition to strong stimulus-pair specificity.