Structure Learning

Updating the structure of the generative model to better account for observations via Bayesian model reduction and expansion.

Neighborhood — ranked by edge-count

framework

Minimum Description Length
associated_with
Related to variational free energy; compressibility corresponds to complexity reduction in structure learning

method

Bayesian Model Expansion
implements
Adding new states or parameters to the generative model if it increases model evidence, enabling concept learning.

concept

Generative Model
implements
Agent's internal probabilistic model of environment; enables belief inference about hidden states given outcomes.
Table 1: Sources of Uncertainty Scored by Expected Free Energy and the Behaviors Entailed
associated_with
Summary table mapping uncertainty types to free energy formulations and corresponding behaviors

cosine ≥ 0.65 · no typed edge

Entities in the same semantic neighborhood but without a typed relation to this one — candidates for new edges or unrecognized duplicates.

Learningconcept0.826
Inference of parameters encoding contingencies of the world (e.g., likelihood matrix A) at slower timescale than perception.
Structure in representationsconcept0.800
The central question of whether representational geometry implies corresponding computational structure
Structure in computationconcept0.798
The actual computational operations a model performs, which the paper argues need not mirror representational structure
Concept Learningconcept0.794
Acquisition of new concepts by Bayesian model expansion and reduction.
Surreptitious teaching of structureclaim0.784
Epistemic Learningconcept0.779
Learning model parameters through curious, uncertainty-reducing behavior; reducing ignorance about contingencies
data structuresconcept0.778
Conventional programming constructs like variables, arrays; claimed unnecessary for Elephant programs.
Associative Learningconcept0.775
Sentience criterion; capacity occurs even in gene regulatory networks and non-neural morphogenetic agents.