Explicit error coding can mediate gain recalibration in continuous bump attractor networks.

Area of Science:

• Computational neuroscience
• Neural network modeling
• Systems neuroscience

Background:

• Continuous bump attractor networks (CBANs) model neural representation of continuous variables.
• Accurate representation relies on precise integration gain, which the brain can recalibrate.
• Existing models lack biologically plausible recalibration mechanisms.

Purpose of the Study:

• To demonstrate biologically plausible mechanisms for integration gain recalibration in ring-type CBANs.
• To explore how established plasticity rules can enable adaptive coding.

Main Methods:

• Instantiated two novel recalibration mechanisms within ring-type CBANs.
• Utilized spatially distributed synaptic plasticity and plasticity in other network components.
• Incorporated explicit error signals to drive plasticity.

Main Results:

• Demonstrated that CBANs can recalibrate integration gain using biologically plausible plasticity.
• Showcased two distinct mechanisms: transient inhomogeneous gain and homogeneous gain recalibration.
• Validated the necessity of error signals for driving plasticity.

Conclusions:

• Ring-type CBANs can achieve adaptive coding of continuous variables through plausible plasticity rules.
• These mechanisms offer a more biologically realistic model for neural computation.
• Provides a foundation for understanding how neural circuits adapt and maintain representations.