Algorithmic Dense Model Theorems and Weak Regularity
summary
Claims a tight equivalence between hardcore lemmas, dense model theorems, and weak regularity, and gives algorithmic dense model theorem and weak regularity lemma reductions that are fully constructive via boosting. Presents an explicit decomposition that peels off O(log 1/δ) dense pieces and produces models whose indistinguishability is provably bounded, shows boosting choices control K versus complexity, and adapts the framework to give a self-reflective GAN with approximate maximum-entropy guarantee plus a complexity-theoretic characterization of computational entropy.
AI-written summary · about