基于智力探索的认知神经动力学

Cognitive Neurodynamics Based on Mental Exploration

啮齿类动物在空间环境中的运动被认为可以形成代表空间环境的认知地图。在这种认知地图中,代表不同空间位置的细胞会与一种发放率编码的模式相联系,这种基于发放率编码的模式同时也广泛地应用于联想记忆的研究中。本文通过神经计算模型和认知地图的构建,阐释了两种认知地图的形成方式和功能。第一种模型是一种空间向量地图,在这种空间向量地图中,动物依据所处的空间位置进行自我定位,同时也可以更新认知地图的信息。另一种认知地图是一个基于目标的向量图,用来进行寻路。将这两种认知地图结合起来,就能形成一个高效快捷的寻路方式。本文重点将这种寻路方法应用于智力探索模型中去,依靠适应性驱动力,利用第一种认知地图中所存地点之间关系的基本知识,对第二种认知地图进行构建和重置。这样通过智力来解决一系列新路径问题的解决,能够针对不同目标的需求找到合适的路径,并且在极短的时间内完成最优路径的形成。这个新思路能为寻路问题找到更加简便和安全的答案,也能够为以后应用于机器人系统的编程提供更好的理论基础。

The spatial motion of rodent hippocampi is considered as a cognitive map to represent a spatial environment. In this cognitive map, different spatial regions are assigned to different cell populations in a framework of rate coding, they are also widely used in the associative memory. Here, the formation and function of the two kinds of cognitive maps were presented through building the neural computational model and cognitive map. The first one was the spatial vector map, it could perform self localization. At the same time, it could update the new detail information of the map. The other one was the goal-oriented vector map, it played an important role in path finding. As an intermediate between the two types, their combination formed an effective and efficient way of path finding. Here, applying this kind of cognitive map-based path-finding method to a mental exploration model was focused on. Relying on the driving force of adaptation, the basic knowledge of the relationship between the locations stored in the first cognitive map was used to form and reset the second cognitive map. Finally, a series of problems of finding new paths were solved with the help of mental exploration to find a suitable path for the needs of different targets within a short time. The fresh idea here could find a way to seek a simpler and safer answer of path-finding, and also provide a better basis for later applying it to the programming of a robot system.