§4.4, p. 353 — the Gentle AdaBoost update is f_m(x) = P_w(y = 1|x) − P_w(y = −1|x)
ProvedAddLogReg.Gentle.update_eq_wprob_diffboostinggentle-adaboostp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a probability measure on , the joint law of , let , fix , and let . The population Gentle AdaBoost increment is the difference of the weighted class probabilities:
where .
This is how the paper contrasts Gentle AdaBoost with Real AdaBoost, whose increment is half the log-ratio of the same two weighted probabilities.
Preamble
import Mathlib import Definitions.Def_AddLogReg_Gentle_Setting open MeasureTheory ProbabilityTheory
Formal statement
namespace AddLogReg.Gentle
theorem update_eq_wprob_diff {X : Type*} [MeasurableSpace X] (ν : Measure (X × Bool))
[IsProbabilityMeasure ν] (F : X → ℝ) (x : X) :
AddLogReg.ExpCrit.wCondExp ν F (fun _ b => AddLogReg.ExpCrit.sgn b) x = wProb ν F true x - wProb ν F false x := by sorry
end AddLogReg.Gentle
Source
Friedman, Hastie and Tibshirani, Additive logistic regression: a statistical view of boosting, Ann. Statist. 28 (2000), pp. 353–354, §4.4, paragraph after the Derivation of Result 4
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.