Reconstructing Neural Network Topology from Firing Activity
The sophisticated topology of the neural network is key to executing normal physiological activity in the process of biological evolution. The reverse engineering of functional connection through populational firing activity is a critical orientation in neuroscience. In this paper, we propose a framework to reconstruct the neural network based on firing activity, which combines the basic principle of gradient descent and cross-correlation analysis. Our framework involves inferring the connection between a pair of neurons and constructing the complete biologically realistic spiking recurrent neural networks. The results suggest that the algorithm is feasible and effective for neural network reconstruction. The reconstructed network and the original network have the same network architecture and firing activity. The framework with the inference algorithm and analysis method can infer the topology of the biological neural network, and improve the research of dynamics underlying neural computations and operational mechanism of the neural network.