Takagi Kosuke
Independent Researcher, Saitama, Japan.
Front Comput Neurosci. 2020 Jan 9;13:86. doi: 10.3389/fncom.2019.00086. eCollection 2019.
Successive patterns of activation and deactivation in local areas of the brain indicate the mechanisms of information processing in the brain. It is possible that this process can be optimized by principles, such as the maximization of mutual information and the minimization of energy consumption. In the present paper, I showed evidence for this argument by demonstrating the correlation among mutual information, the energy of the activation, and the activation patterns. Modeling the information processing based on the functional connectome datasets of the human brain, I simulated information transfer in this network structure. Evaluating the statistical quantities of the different network states, I clarified the correlation between them. First, I showed that mutual information and network energy have a close relationship, and that the values are maximized and minimized around a same network state. This implies that there is an optimal network state in the brain that is organized according to the principles regarding mutual information and energy. On the other hand, the evaluation of the network structure revealed that the characteristic network structure known as the criticality also emerges around this state. These results imply that the characteristic features of the functional network are also affected strongly by these principles. To assess the functional aspects of this state, I investigated the output activation patterns in response to random input stimuli. Measuring the redundancy of the responses in terms of the number of overlapping activation patterns, the results indicate that there is a negative correlation between mutual information and the redundancy in the patterns, suggesting that there is a trade-off between communication efficiency and robustness due to redundancy, and the principles of mutual information and network energy are important to network formation and its function in the human brain.
大脑局部区域激活和失活的连续模式表明了大脑中信息处理的机制。这一过程有可能通过诸如互信息最大化和能量消耗最小化等原则进行优化。在本文中，我通过展示互信息、激活能量和激活模式之间的相关性，为这一论点提供了证据。基于人类大脑的功能连接组数据集对信息处理进行建模，我模拟了该网络结构中的信息传递。通过评估不同网络状态的统计量，我阐明了它们之间的相关性。首先，我表明互信息与网络能量密切相关，并且在同一网络状态附近，这些值分别达到最大化和最小化。这意味着大脑中存在一个根据互信息和能量相关原则组织起来的最优网络状态。另一方面，对网络结构的评估显示，被称为临界性的特征网络结构也在该状态附近出现。这些结果意味着功能网络的特征也受到这些原则的强烈影响。为了评估该状态的功能方面，我研究了对随机输入刺激的输出激活模式。通过测量重叠激活模式数量方面的响应冗余度，结果表明互信息与模式冗余度之间存在负相关，这表明由于冗余，在通信效率和鲁棒性之间存在权衡，并且互信息和网络能量原则对人类大脑中的网络形成及其功能很重要。
