BanditAlgorithm.bernoulli_relative_entropy_pinsker
Provedinequalitiesinformation-theory
(Pinsker's inequality for Bernoulli distributions, L&S Lemma 10.2(b)) For and :
(The book states Lemma 10.2 for ; at the inequality is the contentless , resp. , which the real-valued junk convention cannot express, hence . Lemma 10.2's other parts — convexity of and with unique minimizers at resp. , and part (c): — are recorded in the Lemma 10.2 milestone.)
Preamble
import Definitions.Def_bernoulliRelativeEntropy
Formal statement
theorem BanditAlgorithm.bernoulli_relative_entropy_pinsker (p q : ℝ)
(hp : p ∈ Set.Icc (0 : ℝ) 1) (hq : q ∈ Set.Ioo (0 : ℝ) 1) :
2 * (p - q) ^ 2 ≤ bernoulliRelativeEntropy p q := by
sorry
Source
L&S Lemma 10.2, p.135
Read-back
What the Lean code literally says, in plain math · claude-fable-5
Setup and notation. Let denote the function
on pairs of reals, computed with the conventions that division by yields and of any nonpositive number is (so is real-valued everywhere).
Assertion. For all real ,
Hypotheses.
- (closed interval; both endpoints allowed).
- (open interval; both inequalities strict).
Edge cases.
- At the first term is (by multiplication with zero), so ; at , .
- Because is strictly interior, the conventional finite boundary values and play no role.
- The conclusion is a non-strict inequality.
Human review
Confirmed by the mission captain (proposal self-audit).