Information theoretic analysis of energy efficient neurons with biologically plausible constraints

In this paper we investigate the consequences of biologically plausible constraints on predictions of the Berger-Levy energy efficient neuron model. As new constraints we consider i) a peak power constraint, ii) peak energy expenditure per ISI constraint, iii) a lower bound on the value of inter spike interval (ISI), and iv) lower and upper bounds on the excitatory postsynaptic potential (EPSP) intensity, λ. Our analysis shows that considering these constraints of the capacity per unit cost maximization problem changes the shape of probability distribution function (PDF) of λ and the ISIs. We show, using numerical solutions of the optimization problem, that the new constraints change the PDFs of λ and the ISIs in term of their shape and location of the peak value. We also derive predictions for how the coefficient of variation (CV) of the ISI is changed, which is easier to characterize experimentally than the full PDF.