Bayesian efficient coding as a theory of perception: progress, controversies, and prospects.

Area of Science:

• Cognitive Neuroscience
• Computational Neuroscience
• Sensory Perception

Background:

• Efficient coding and Bayesian inference are key theories of sensory processing.
• Natural environmental statistics are hypothesized to shape sensory encoding and interpretation.
• Previous models did not fully integrate these two theories.

Purpose of the Study:

• To review the Bayesian efficient coding framework, which unifies efficient coding and Bayesian inference.
• To examine behavioral and neural evidence supporting this theory in perception and cognition.
• To discuss theoretical extensions and their implications for understanding sensory adaptation and biases.

Main Methods:

• Review of existing behavioral and neural evidence.
• Analysis of theoretical developments in Bayesian efficient coding.
• Examination of adaptation mechanisms within the framework.

Main Results:

• The Bayesian efficient coding framework successfully explains perceptual biases and discrimination thresholds.
• Environmental statistics are shown to lawfully relate to bias and discrimination.
• Adaptation, both short- and long-term, can be modeled within this framework.
• Response biases can be decomposed into encoding and decoding components.

Conclusions:

• Bayesian efficient coding provides a unified account of sensory processing.
• The theory has evolved significantly over the past decade, addressing limitations.
• It offers a powerful framework for future research in perception and cognition.