The motivations of this work are the comprehensive construction and computational analysis of a model structure derived objectively from experimental findings in brain neural data, including the observation that an unsupervised decoded trajectory exhibits a degree of symmetry with the actual trajectory within the activity space. In this study, we are also inspired by the formation of grid cells to create a more general and robust grid module and to construct an interactive, self-reinforcing learning system that incorporates Bayesian inference; the proposed approach is interpreted as spatial division and exploration-exploitation with grid feedback, abbreviviated as Grid-SD2E. Here, the grid module can be used as an interaction medium between the outside world and a system, as well as a self-reinforcement medium within the system. The spatial division and exploration-exploitation (SD2E) mechanism receives the 0/1 signals of a grid through its spatial division (SD) module. Herein, we analyse the rationality of the target system on the basis of existing theories in both neuroscience and cognitive science and propose a general learning principle (i.e., special and general rules) to explain the different interactions between people and between people and the external world. Finally, we believe that the Grid-SD2E framework should be regarded as a computational model of the brain (essentially a cognitive learning system). However, it should be noted that since the model structure is derived akin to a geometric drawing, making its scientific implications difficult to understand, we attempt to interpret this model structure from multiple perspectives and disciplines rather than venturing into other fields.
Feng et al. (Wed,) studied this question.