Free-lunch learning and directional distributions in artificial neural networks
Keywords:Grassmann manifold, isotropy, sphere, uniform distribution
Free-lunch learning (FLL) is a phenomenon in which relearning partially-forgotten mental associations induces recovery of other
associations. FLL occurs also in artificial neural networks (ANN’s). Two models for forgetting in ANN’s are presented which involve uniform distributions on spheres and Grassmann manifolds. It is shown that these models differ markedly in their amounts of FLL.