A Neural Network Model of Self-Organized Learning in Hippocampus- Entorhinal Cortex Spatial Representations

In this paper, a novel self-organizing neural network model is proposed for navigation. The model, consisting of two modules, learns to encode direction and position information based on both proximal and distal landmarks in the environment while integrating self-motion derived from visual information. This multimodal approach allows for more accurate self-motion encoding and improved model robustness. One of the key contributions of this model is that it is fully self-organizing in a biologically plausible manner. All functionally relevant connectivity is plastic and subject to local associative Hebbian learning, breaking through the limitation of non-plasticity connections often assumed in earlier attractor network models. Additionally, the model utilizes the neural circuit mechanism of head orientation cells and place cells to accomplish spatial coding. The results of this study demonstrate the potential of extending the model to solve autonomous navigation problems in complex environments. Overall, the proposed model offers an improvement over existing brain-inspired navigation algorithms, providing a promising direction for future research in brain-inspired navigation systems.