Department of Computational and Applied Mathematics

The brain encodes and stores spatial information via spatially tuned neurons known as place cells. I approximate neural response to exploration of a novel environment using a dynamic network model of isopotential place cells with plastic connections. In this talk, I propose a neural readoff mechanism for path integration, described by an algorithm developed from persistent homology theory. This algorithm provides a measure of efficiency in the network model by computing the minimum exploration time necessary to reconstruct topological features of the environment from neural activity. The minimum time offers a theoretical lower bound for the amount of exploration time required to learn a new environment.