Bayesian inference is a statistical framework where the brain combines prior belief which incoming sensory evidence the most probable of sensation.Prior represent expectation based on past experience, while sensory data new evidence.Predictive coding is a neural implementation of Bayesian inference.
How do they work:
- Hierarchical processing: Predictive coding organizes the brain into layers with higher level generating prediction about lower-level sensory output. Bayesian Inference provide the mathematical framework of updating this prediction based on the likelihood of observed data.
2.Error minimization: Predictive error act as the signal for Bayesian updating. The brain adjusts it previous to reduce future error, effectively performing approximate Bayesian inference in real time.
KEY DIFFERENCES BETWEEN PREDICTIVE CODING AND BAYESIAN INFERENCE
. Level of Abstraction: Bayesian inference in a computation principal describing optimal probalistic reasoning while predictive coding is a neural mechanism that implement this principal in the brain.
. Focus: Bayesian Inference emphasizes the combination of prior and likelihood mathematically, whereas predictive coding emphasizes error signaling and in neurocircuits.
APPLICATION
Predictive coding explains phenomena like hallucinations or sensory deficits as failure deficits as failure of Bayesian updating, where prior is mispositioned or predictive error are improperly weighted.
In summary, predictive coding operationalizes Bayesian Inference in the brain, using hierarchical prediction and error signals to update belief and action while Bayesian Inference provides the theoretical foundation, predictive coding provides a biologically plausible mechanism for implementing these computation and circuits.
SOURCES:
With or without you: predictive coding and Bayesian inference in the brain
