Coarse lower Mills bound for the standard Gaussian tail
ProvedBanditAlgorithm.standardGaussian_mills_lower_boundanalysisgaussianmills-ratioprobability
There is a universal constant such that the upper tail of a standard Gaussian satisfies, for every ,
This coarse lower Mills bound is obtained by integrating the Gaussian density over the short interval .
Preamble
import Mathlib.Probability.Distributions.Gaussian.Real open MeasureTheory ProbabilityTheory
Formal statement
namespace BanditAlgorithm
/-- A coarse Mills-ratio lower bound sufficient for Thompson sampling. -/
theorem standardGaussian_mills_lower_bound :
∃ c : ℝ, 0 < c ∧
∀ u : ℝ, 0 ≤ u →
c / (u + 1) * Real.exp (-u ^ 2 / 2) ≤
(gaussianReal 0 1).real (Set.Ioi u) := by
sorry
end BanditAlgorithmSource
Lattimore and Szepesvári, Bandit Algorithms (CUP 2020), Gaussian-tail hint to Exercise 36.6, printed p. 475 / PDF p. 484: the displayed lower bound on the standard normal upper tail.