Figure 5.

Parameters of a WED analysis can be altered to reproduce biologically realistic constraints on decoding. The different neurons can be kept as separate dimensions in Euclidean space (‘N: separated’) or combined (i.e., average population response) mimicking the fact that synaptic inputs could combine in the postsynaptic terminals of decoding neurons before further processing (‘N: combined’). Weights are calculated independently for each time bin (see Figure 4) and averaged across time to be kept fixed for a given neuron (‘W: fixed’). Alternatively, weights can be fitted to follow a specific function (‘W: trend’). Examples of functions that could be implemented include rules that would replicate firing rate-based plasticity such as facilitation and depression. A firing rate-dependent change in weight across time was not beneficial in the case of the three model systems examined here: (A). Electrosensory; (B). Olfactory; (C) LIF model. Curves show averages (± s.d.) across all pairs of stimuli.