C and D elucidate the impact of the plateau potentials probability f q and of the LTD probability (default value 0.5) on the repulsion effect. The strength of repulsion was measured by the repulsion index, which is obtained by dividing the overlap ratio of memory traces for a pair of unrelated memory items by the overlap ratio for a pair of memory items with 40% overlap. Values above the red dashed lines that are defined by value 1 indicate the occurrence of the repulsion effect. One sees that values for f q that are below our biologically derived default value 0.005 do not produce repulsion and also that the LTD component of BTSP is essential for repulsion. In fact, the repulsion effect becomes even stronger when the LTD probability is increased beyond its default value. The error bars in both panels represent the standard deviation of the overlap ratio over 20 trials with new generations of random connection weight, input patterns, and plateau potentials.
