Learning Grid Cells with Vector Representation of Self-Position and Matrix Representation of Self-Motion

• Activity Number: 436 - Deep Learning for Data Science
• Type: Invited
• Date/Time: Wednesday, July 31, 2019 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Ying Nian Wu*
• Companies: UCLA
• Keywords: Grid cells; Representation; Path integral; Path planning; Error correction; Deep learning

Abstract

We propose a mathematical and computational model for learning the grid cells that are observed in the brains of many species for spatial awareness and navigation. In this model, the self-position of the agent is represented by a vector, and the self-motion of the agent is represented by a block-diagonal matrix. The learned units exhibit hexagon grid patterns that characterize the grid cells. The learned model can be used for path integral and path planning. Moreover, the learned representation is capable of error correction. Based on joint work with Ruiqi Gao, Jianwen Xie and Song-Chun Zhu.