Robust design of inhibitory neuronal networks displaying rhythmic activity

Central pattern generators (CPGs) are neuronal circuits which autonomously produce patterns of phase-locked activity. The need for bioelectronic implants that adapt to physiological feedback calls for novel methods
for designing synthetic CPGs that respond identically to their biological
counterparts. Nonlinearity within such networks make both the prediction of network behaviour and, inversely, the design of networks with desired behaviour non-trivial. Here, we demonstrate the utility of optimization-based parameter estimation for identifying sets of synaptic parameters which give rise to network activity with specic temporal properties. We reduce the dimension of the problem by visualizing the network dynamics in a coordinate system composed of the relative phases of the oscillators, and we reduce each oscillator to its phase response curve (PRC). While recent work has employed parameter estimation to optimize single neuron models, reducing the component oscillators to their PRCs allows us to estimate parameters of networks of arbitrarily complex neuron models without incurring prohibitive computational costs. We highlight a possible application of our approach by estimating parameters of a CPG emulating the phase-locked activity associated with ECG data. This work paves the way for the design of synthetic networks which may be interfaced with nervous systems.