A Fast and Efficient Ensemble Transfer Entropy and Applications in Neural Signals.

Abstract

The ensemble transfer entropy () refers to the transfer entropy estimated from an ensemble of realizations. Due to its time-resolved analysis, it is adapted to analyze the dynamic interaction between brain regions. However, in the traditional , multiple sets of surrogate data should be used to construct the null hypothesis distribution, which dramatically increases the computational complexity. To reduce the computational cost, a fast, efficient with a simple statistical test method is proposed here, in which just one set of surrogate data is involved. To validate the improved efficiency, the simulated neural signals are used to compare the characteristics of the novel with those of the traditional . The results show that the time consumption is reduced by two or three magnitudes in the novel . Importantly, the proposed could accurately track the dynamic interaction process and detect the strength and the direction of interaction robustly even in the presence of moderate noises. The novel reaches its steady state with the increased samples, which is slower than the traditional method. Furthermore, the effectiveness of the novel was verified in the actual neural signals. Accordingly, the proposed in this work may provide a suitable way to investigate the dynamic interactions between brain regions.