Title:An optimization method for predicting the functional mode of a biological neuronal network

Abstract:We employ an optimization procedure to estimate parameters of a small model biological neuronal network, which assumes multiple modes of activity, depending on parameter values. The input to the network and resulting network output are: stimulating electrical current delivered to each constituent neuron, and membrane voltage of each neuron, respectively. We write a cost function in terms of a measurement and a model error. An iterative reweighting of these terms permits a systematic method to identify the lowest-minimum, within a local region of state-and-parameter space, on the surface of a non-convex cost function. We use the estimates corresponding to the lowest minimum of the cost function to predict the particular mode of activity that the network is expressing. Finally, we show how this method can reduce a model of unnecessarily high complexity to a representation that contains the maximum dimensionality contained in the available measurements. The design of simulated experiments presented in this paper might apply in an experimental laboratory, to estimate properties of extremely small biological circuits.