Estimation of network structures only from spike sequences

A neuron, the fundamental element of neural systems, interacts with other neurons, often producing very complicated behavior. To analyze, model, or predict such complicated behavior, it is important to understand how neurons are connected as well as how they behave. In this paper, we propose two measures, the spike time metric coefficient and the partial spike time metric coefficient, to estimate the network structure, that is, the topological connectivity between neurons. The proposed measures are based on the spike time metric and partialization analysis. To check the validity, we applied the proposed measures to asynchronous spike sequences that are produced by a mathematical neural network model. It was found that the proposed measure has high performance for estimating the network structures even though the structures have a complex topology such as a small-world structure or a scale-free structure.