Bayesian efficient coding.

Bayesian efficient coding defines optimal neural codes by combining a stimulus prior, encoding model, capacity constraint, and a posterior loss functional, unifying information and non-information objectives.

• Why it matters: The efficient coding and Bayesian Brain hypotheses connect information maximization with Bayesian inference, but a general optimality framework for arbitrary posterior-based objectives has been missing.
• What they did: The authors formalize Bayesian efficient codes using four ingredients and show that classic entropy-based efficient codes are a special case; they introduce covtropy losses and analyze how decorrelation changes under error-penalizing objectives.
• The result: In the blowfly large monopolar cell, Laughlin’s contrast-coding data are better explained by minimizing Lp reconstruction error with p=1/2 than by information maximization, overturning a forty-year-old interpretation.

AI summary of the abstract · may contain errors · not medical advice · read the paper.