Evidence or Marginal? What Moment-Vector Propagation Computes in Polynomial Density Neurons
En palabras de los autores
When a neuron's input is lost, or known through a belief , what should it output? For a neuron storing a joint polynomial density with input density , two formulas compete. costs one matrix-vector product; Hierarchical Correlation Reconstruction introduces it as a constant-denominator approximation to . For a density model both are exact answers to different questions: is Bayesian updating with the belief as virtual evidence; installs the belief as a new marginal (Jeffrey's rule). With a missing input returns the model's conditional; generally does not. This post hoc reading is proved, and illustrated outside the pair family up to inputs; it has no pre-registered test, only one exploratory real-data pilot. Post hoc, closed forms fall within a factor 2 of all 112 measured gaps, and negative density begins near input correlation . In a pre-registered synthetic study (up to 32 inputs), 5 of 15 hypotheses are supported, 2 inconclusive and 8 not supported, including all three on transfer beyond equicorrelated Gaussians. Post hoc, where is defined, truncation usually outweighs the gap for Beta beliefs in pair neurons; triple neurons under flat and data beliefs are exceptions.
Apareció: martes, 10 de noviembre. arXiv. Preprint, todavía sin revisión por pares.
Comentario de los autores: 33 pages, 12 figures, 33 tables. Code: https://github.com/Przemyslaw11/EvidenceOrMarginal