A Gaussian with a linear term is integrable
ProvedBanditAlgorithm.integrable_exp_neg_quadratic_add_linearreference
For and any the function is Lebesgue integrable on .
Mathlib has integrability of the pure Gaussian but not of the completed-square form with a linear term. The two are related by the translation , which is also what evaluates the integral, so this lemma is the integrability half of the Gaussian-with-linear-term package and is what lets the method-of-mixtures computation be transported from Bochner integrals to -valued ones, where Tonelli's theorem applies.
Preamble
import Mathlib.MeasureTheory.Measure.Prod import Mathlib.Analysis.SpecialFunctions.Gaussian.GaussianIntegral open MeasureTheory Real
Formal statement
theorem BanditAlgorithm.integrable_exp_neg_quadratic_add_linear {a : ℝ} (ha : 0 < a)
(c : ℝ) :
MeasureTheory.Integrable (fun x : ℝ ↦ Real.exp (-a * x ^ 2 + c * x)) := by
sorrySource
Standard: the integrability companion to the completed-square Gaussian integral. Needed to transport the mixture computation from Bochner integrals to [0, infinity]-valued ones.