Figure 2.

Obtaining the compensating synaptic kernels. The compensation kernel η(t) was obtained by stimulating a single model neuron with a fluctuating (OU) current and keeping track of the times t_j that the voltage crossed a threshold from below (black dashed line in first panel). For each spike, we then obtain an action potential waveform V_AP^j(t) for t_j ≤ t < t_j + t_s, where t_s will set the width of the η(t) kernel (we take t_s = 4 ms). We then sum these traces to obtain the average waveform of the action potential V_AP(t) (black dashed line in the second panel). The kernel η(t) is then the temporal derivative of this averaged action potential waveform (third panel). η(t) represents an approximation to the total change in voltage of the neuron during an action potential. Last, η(t) is convolved with an exponential function to obtain the synaptic kernel η̃(t) (last panel). Note that η̃(t) changes sign but also very rapidly goes to zero as time goes on. Inset in the last panel shows the η̃ kernel over a shorter period of time.