6 Supplementary material All spiking network simulations were run using Auryn [

Network simulations performed in fSBI were distributed over 300 cores from four workstations for two months. The resulting 1D signal was Fourier-transformed. Fano factor was computed per neuron, then averaged over windows. We also do not correct for the change in prior at each round. We here clarify the relationship between the pseudo-posterior obtained from fSBI and the corresponding true Bayesian posterior. New metrics only at each round: We recall from Eqns. 10 that after two rounds of filtering, fSBI Note that the true Bayesian posterior after two rounds would be, using Bayes' rule: p The diagonals show the distribution of values for each plasticity parameter.