Text S1 - Predictive Coding of Dynamical Variables in Balanced Spiking Networks

In the supporting Text S1, we address several issues regarding the biological plausibility of our approach and the generality of our results. In particular, we demonstrate that the networks can be separated into purely excitatory and inhibitory neurons. We discuss the problem of perturbing synaptic weights and show that the networks are robust to synaptic failures and to noise in the lateral connections. We furthermore derive analytical results for the dynamics of a network with identical kernels tracking a scalar dynamical variable, , in the limit of high firing rates.
(PDF)