| Popis: |
The entorhinal cortex uses grid-like neural representations (grid cells) for spatial processing. Neural representations for concepts (concept cells), which presumably support high-level conceptual processing, have been also found in the same brain region. However, the relationship behind those neural representations is still unclear. We propose a unified model called “disentangled successor information (DSI)” that produces neural representations for physical space and linguistic concepts. DSI generates gridlike representations in a 2-dimensional space that highly resemble those observed in the brain. Moreover, the same model creates concept-specific representations from linguistic inputs, corresponding to concept cells. Mathematically, DSI approximate value functions for navigation and word vectors obtained by word embedding methods, thus those representations support both spatial navigation and analogical inference based on a simple arithmetic computation. Our results suggest that representations and computation for physical space and linguistic concepts in the brain can emerge from a shared mechanism. |
