Free energy and learning
Using the free energy principle to derive multiple theories of associative learning allows us to combine them into a single, unifying framework.
Associative learning, information and active inference
Submitted to Neural Computation (2023)
Associative learning, information and active inference In this paper we demonstrate that the free energy principle can explain and combine into a single framework the principles of associative learning, derived earlier in the studies of animal behavior in conditioning experiments, and make some assumptions regarding the mechanisms of learning. However, only a few experiments have been modeled, and the literature on associative learning is vast. Thus, the goal of further research may be to model the described experiments on animals and humans, as well as to plan new ones that test the principles of learning as free energy minimization.
Submitted to Neural Computation (2023)