Urgent.News

What's breaking now, across thousands of outlets.

AI

The global geometry of optimal neural representations

Efficient coding hypothesizes that neural activity optimally represents information under resource constraints. The global structures of such optimal population codes are difficult to study theoretically, and previous works have usually resorted to either focusing on local properties like fine discrimination or placing restrictions on the parametric forms of neural tuning. We address the general…

The efficient coding hypothesis posits that neural activity represents information optimally while adhering to resource constraints. Traditional approaches to studying optimal population codes have focused on local properties or imposed restrictions on neural tuning, making the analysis of global structures challenging. This research aims to address the general problem by deriving the neural representation that minimizes the all-pairs average classification error for a one-dimensional circular variable under various forms of neural response noise.

The approach involves viewing the space of neural response distributions as a statistical manifold, which can be simplified using a generalized version of variance-stabilizing transformations.

The theory developed in this study fully characterizes the entire family of optimal population codes and enables the formulation of two novel predictions. Firstly, it suggests that an optimal code should permit the coexistence of oppositely peaked bimodal tuning curves. Secondly, it posits a correlation between spike count variability and tuning sharpness. Both predictions were validated in the context of the mouse head direction system.

Written by urgent.news from bioRxiv's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.

Read the original at biorxiv.org →

More in AI

More from Wednesday 18 November →