Bernoulli model: an event and its complement have total mass one
Provedbernoulli_event_complement_probabilityReusable Bernoulli foundational lemma. For any inclusion probability and any event on the observation set, the Bernoulli-model probabilities of and of its complement sum to one: , equivalently . This is the total-mass normalization (here ) underlying every concentration complement argument in the matrix-completion tree. Note no hypothesis is needed: the identity is the algebraic Fintype product expansion .
Preamble
import Definitions.Def_matrix_completion_bernoulli open MatrixCompletion
Formal statement
theorem bernoulli_event_complement_probability
{n₁ n₂ : ℕ} (p : ℝ) (E : Finset (Fin n₁ × Fin n₂) → Prop) :
bernoulliEventProb p E + bernoulliEventProb p (fun Ω => ¬ E Ω) = 1 := by
sorrySource
Elementary product-Bernoulli normalization (Fintype.prod_add); foundational complement identity for the Bernoulli sampling model of Candes-Recht, CACM 55(6):111-119, 2012.