基于完全图的脑功能时序网络模式及其能量动力学研究

Revealing the human brain functional cognition would provide an opportunity to break the bottleneck of current artificial intelligence developments. Brain functional cognitive dynamic network modelling is one of very important approaches. In this project, we explore the brain functional time series network model based on the complete graphs and its energy dynamics. We transcribe brain functional activities into the multi-voxel time series with frequency, phase and amplitude. Using the frequency and phase of time series, we present a features extraction algorithm from brain functional activities to avoid the false positive and false negative in statistical tests. According to the dense trait of the network topology model, the brain functional network reconstruction approach based on complete-graph skeleton is proposed to solve the non-robust problems caused by noise and non-equilibrium. And then we investigate the dynamics about the brain functional networks through the phase and energy spectrum of its time series, in which the evolution of the networks are deduced by the phases of time series, and the energy fields are derived from its amplitudes. We will also depict the cognitive states and evolution processes in human brain language cognition contrasting Chinese and English to simulate the changes and transitions of energy fields and conscious streams with different periods. This research work will provide important approaches to understand the brain functional language cognition and to analysis the mind and intelligence of human beings.

人工智能发展的突破口在于揭示人类认知，脑功能认知动态网络是重要途径之一。本项目研究基于完全图的脑功能时序网络模式及其能量动力学，将脑功能活动转录为频率、相位、幅度三位一体的多体素化时间序列，利用时间序列频率和相位，提出基于体素化时间序列动态弹性匹配的脑功能特征抽取方法，避免统计检验中脑激活特征提取出现的假阳性和假阴性；依网络拓扑模式稠密性状，提出基于完全图骨架的脑功能网络重构理论和方法，解决噪声和非平衡带来的非鲁棒性难题；提出基于时间序列相位和能量谱的脑功能网络动力学分析方法，通过时间序列相位拓展时间效应演绎网络演化，从时间序列幅度导出能量谱探索演化驱动力；比对中英语言认知，模拟脑功能语言认知能量场和意识流动态变迁，利用可视化网络来展示人脑语言认知活动和演化，有望为人类语言认知以及依语言外在表征的思维和智能解析提供新方法。

人工智能发展的突破口在于揭示人类认知，脑功能认知动态网络是重要途径之一。本项目研究基于完全图的脑功能时序网络模式及其能量动力学，将脑功能活动转录为频率、相位、幅度三位一体的多体素化时间序列，利用时间序列频率和相位，提出基于体素化时间序列动态弹性匹配的脑功能特征抽取方法，避免统计检验中脑激活特征提取出现的假阳性和假阴性；依网络拓扑模式稠密性状，提出基于完全图骨架的脑功能网络重构理论和方法，解决噪声和非平衡带来的非鲁棒性难题；提出基于时间序列相位和能量谱的脑功能网络动力学分析方法，通过时间序列相位拓展时间效应演绎网络演化，从时间序列幅度导出能量谱探索演化驱动力；比对中英语言认知，模拟脑功能语言认知能量场和意识流动态变迁，利用可视化网络来展示人脑语言认知活动和演化，有望为人类语言认知以及依语言外在表征的思维和智能解析提供新方法。