Initialize fan-in-normalized weights. For an I→E matrix with 1024 excitatory columns and ${𝑁}_{𝐼}$ inhibitory rows, each weight was ${𝑊}_{𝑘𝑗}^{𝐼𝐸}=\frac{{𝐺}_{\text{draw}}}{{𝑁}_{𝐼}}$, with ${𝐺}_{\text{draw}}=max\left(0,𝑋\right)$ and $𝑋\sim \mathrm{𝒩︀}\left({𝜇}_{\text{init}},{𝜎}_{\text{init}}^2\right)$ a Gaussian draw of mean ${𝜇}_{\text{init}}$ and standard deviation ${𝜎}_{\text{init}}=0.1{𝜇}_{\text{init}}$, in μS. Defining expected summed coupling ${𝐺}_{𝐼\to 𝐸}=\mathrm{ℰ︀}\left[{𝐺}_{\text{draw}}\right]$ gives

Here $\mathrm{ℰ︀}$ averages over weight initialization, $𝑘$ indexes inhibitory neurons and $𝑗$ excitatory neurons; ${\overline{𝑤}}_{𝐼\to 𝐸}$ is the expected conductance of one synapse, not an identical realised weight. An inhibitory volley gives

where ${\mathrm{ℐ︀}}_{\text{active}}$ is the active inhibitory set and $\mathrm{\Delta }{𝑔}_{\text{inh,post=E},𝑗}$ the inhibitory conductance increment at E neuron $𝑗$, in μS; both weights and participation therefore enter the increment.

Apply paired pool-size controls. We swept ${𝑁}_{𝐼}\in \left\{16,64,256\right\}$. Fixed-summed controls used parent means ${𝜇}_{\text{init}}\in \left\{1,2,4\right\}$ μS, so ${\overline{𝑤}}_{𝐼\to 𝐸}\propto \frac{1}{{𝑁}_{𝐼}}$; fixed-mean-synapse controls scaled ${𝜇}_{\text{init}}$ with ${𝑁}_{𝐼}$, giving nominal ${\overline{𝑤}}_{𝐼\to 𝐸}\in \left\{3.90625,7.8125,15.625\right\}$ nS. Nominal values approximate the post-clamp expectations; the arms coincide at 256 I neurons, with one additional shared condition at 64 I neurons and nominal 1 μS summed coupling.