基于神经能量场梯度的智力探索

发布时间：2019-01-07 12:29

【摘要】：啮齿类动物可以在海马中形成代表环境的认知地图来解决空间问题。在经典的认知地图模型中,系统需要大量物理探索来学习空间环境,从而解决寻路问题,这个过程会耗费大量时间和能量。虽然Hopfield的智力探索模型弥补了这个缺陷,但是该模型并没有关注路径的高效性。并且这个模型主要来自于人工神经网络,缺乏明确的生理学意义。该计算模型在智力探索概念的基础上,运用神经能量编码的理论来解决路径搜索问题:该模型通过位置细胞集群的发放功率构建一种能量场,并计算能量场的梯度,进而用梯度向量来研究智力探索问题。研究结果表明提出的这种新的智力探索模型不仅可以更高效地找到优化路径,而且呈现了具有生物物理学意义的学习过程。这种新思路验证了位置细胞和突触对空间记忆的重要性以及能量编码的有效性,为了解空间记忆的神经动力学机制提供了理论基础。
[Abstract]:Rodents can form cognitive maps representing the environment in the hippocampus to solve spatial problems. In the classical cognitive map model, the system needs a lot of physical exploration to study the spatial environment, thus solving the problem of finding the way. This process will consume a lot of time and energy. Although Hopfield's intelligence exploration model makes up for this flaw, the model does not focus on the efficiency of the path. And this model mainly comes from artificial neural network, lack of clear physiological significance. Based on the concept of intelligence exploration, this model uses the theory of neural energy coding to solve the path search problem: the model constructs an energy field through the power release of the location cell cluster, and calculates the gradient of the energy field. Then the gradient vector is used to study the problem of intelligence exploration. The results show that the proposed new intelligence exploration model can not only find the optimal path more efficiently, but also present a biophysical learning process. This new idea verifies the importance of location cells and synapses to spatial memory and the validity of energy coding and provides a theoretical basis for understanding the neurodynamic mechanism of spatial memory.
【作者单位】：华东理工大学认知神经动力学研究所;
【基金】：国家自然科学基金(11232005;11472104) 教育部博士生基金(20120074110020)
【分类号】：TP183;Q42

【相似文献】