Theory One: Successor Representation 

This piece, by Armin Bazarjani, was published on 11/19/24. The original text, by Stachenfeld et al., was published by Nature Neuroscience on 10/02/17.

Prior to the publication of this paper, it was formalized that, through place cells and grid cells, the hippocampus most likely played some sort of role in constructing our internal representation of the world around us, a cognitive map. This paper (Stachenfeld et al. 2017) proposed that the hippocampus may encode something more akin to a predictive map.

The authors suggest that the hippocampus encodes a “successor representation” (SR,) which encodes the expected future occupancy of different states. This challenged the traditional view, at the time, of a purely spatial cognitive map as they were unable to explain many different hippocampal phenomena, such as the dependence of place fields on the animal’s behavioral policy and environmental topology.

They showed that the SR can explain many properties of place cells, like those listed above. The authors also proposed that grid cells in the entorhinal cortex encode a compressed, low-dimensional basis set for the SR, which is useful for regularization and hierarchical planning.

Stachenfeld et al. formalized the notion of predictive function into a reinforcement learning framework. By doing so, they were also able to give a mechanistic account of why place cells and grid cells fire the way they do and how they are encoded mathematically within this larger framework of a predictive map.

References

Stachenfeld, K. L., Botvinick, M. M., & Gershman, S. J. (2017). The hippocampus as a predictive map. Nature neuroscience, 20(11), 1643-1653.

Want to submit a piece? Or trying to write a piece and struggling? Check out the guides here!

Thank you for reading. Reminder: Byte Sized is open to everyone! Feel free to submit your piece. Please read the guides first though.

All submissions to berkan@usc.edu with the header “Byte Sized Submission” in Word Doc format please. Thank you!