Combining sampling and attractor dynamics in spiking models of head direction systems

Area of Science:

• Neuroscience
• Computational Neuroscience
• Systems Neuroscience

Background:

• Neural circuits must integrate sensory information despite inherent uncertainty.
• Attractor-based neural circuits, like those for head-direction (HD) representation, show cue-dependent precision.
• Understanding how neural dynamics achieve stability during uncertain computation is crucial.

Purpose of the Study:

• To propose a spiking neural network model that integrates stability and uncertainty.
• To reinterpret the head-direction (HD) circuit as an uncertainty-aware integrator.
• To explain how neural populations compute under uncertainty.

Main Methods:

• Developed a spiking neural network model unifying attractor dynamics and probabilistic inference.
• Modeled neural population activity representing uncertainty through rapid fluctuations.
• Simulated the network to generate predictions for experimental validation.

Main Results:

• The model demonstrates how neural populations can maintain stable representations while fluctuating to encode uncertainty.
• Head-direction (HD) bump precision is shown to decrease due to fluctuations reflecting input uncertainty.
• Generated experimentally testable predictions including correlated voltage fluctuations and specific bump movement statistics.

Conclusions:

• A unified framework combining probabilistic inference and attractor dynamics explains uncertainty-aware computation in neural circuits.
• Neural populations can represent estimates and their uncertainty via fluctuations while maintaining stability.
• This principle may be generalizable to various noisy biological systems for robust information processing.