Associative Learning and Active Inference.

Area of Science:

• Neuroscience
• Cognitive Science
• Computational Psychiatry

Background:

• Associative learning connects stimuli based on co-occurrence, foundational to understanding behavior.
• Classical models like Rescorla-Wagner explain learning via reward prediction errors but lack scope.
• Existing models struggle to encompass the full spectrum of learning phenomena.

Purpose of the Study:

• To apply the free energy principle to associative learning, framing it as uncertainty minimization.
• To link the free energy principle with the Rescorla-Wagner model and explore informational aspects of learning.
• To model behavioral phenomena like blocking, overshadowing, and latent inhibition within the active inference framework.

Main Methods:

• Utilizing the free energy principle to model learning as surprise minimization.
• Investigating informational aspects, surprise types, and prediction errors.
• Applying the active inference framework to explain behavioral phenomena and attention.

Main Results:

• The free energy principle provides a unified framework for associative learning.
• Behavioral phenomena are modeled by integrating informational and novelty aspects of attention.
• Demonstrates a link between free energy minimization and prediction error-based learning.

Conclusions:

• The free energy principle offers a comprehensive theoretical basis for associative learning.
• It integrates diverse empirical models and explains complex behavioral phenomena.
• Provides a framework for understanding the brain's computational processes in learning.