Documentation

ZPM.Probability.Decision.Minimax.MWU.Run

Running MWU for T rounds and the T-step potential bound #

mwuRun iterates mwuUpdateWeights for T rounds, threading the best-response selector produced by the hrow hypothesis. The T-step potential bound Φ_T ≤ |C| · (1 - η · v)^T follows by induction from the one-step bound.

def ProbabilityTheory.mwuRun {R : Type u_1} {C : Type u_2} [Fintype R] [Fintype C] [Nonempty C] (M : RCBool) (η : ) (hη1 : η < 1) (v : ) (hrow : ∀ (q : FintypePMF C), ∃ (r : R), v c : C, q.prob c * if M r c = true then 1 else 0) (T : ) :
MWUConfig C × (Fin TR)

Run MWU for T rounds: returns the final configuration and the row sequence.

Equations
Instances For
    @[reducible, inline]
    noncomputable abbrev ProbabilityTheory.mwuConfig {R : Type u_1} {C : Type u_2} [Fintype R] [Fintype C] [Nonempty C] (M : RCBool) (η : ) (hη1 : η < 1) (v : ) (hrow : ∀ (q : FintypePMF C), ∃ (r : R), v c : C, q.prob c * if M r c = true then 1 else 0) (T : ) :

    The MWU configuration after T rounds.

    Equations
    Instances For
      @[reducible, inline]
      noncomputable abbrev ProbabilityTheory.mwuRows {R : Type u_1} {C : Type u_2} [Fintype R] [Fintype C] [Nonempty C] (M : RCBool) (η : ) (hη1 : η < 1) (v : ) (hrow : ∀ (q : FintypePMF C), ∃ (r : R), v c : C, q.prob c * if M r c = true then 1 else 0) (T : ) :
      Fin TR

      The MWU row sequence after T rounds.

      Equations
      Instances For
        theorem ProbabilityTheory.mwu_potential_T_bound {R : Type u_1} {C : Type u_2} [Fintype R] [Fintype C] [Nonempty C] (M : RCBool) (η : ) ( : 0 η) (hη1 : η < 1) (v : ) (hrow : ∀ (q : FintypePMF C), ∃ (r : R), v c : C, q.prob c * if M r c = true then 1 else 0) (T : ) :
        (mwuConfig M η hη1 v hrow T).potential (Fintype.card C) * (1 - η * v) ^ T

        T-step MWU potential bound: after T rounds, Φ_T ≤ |C| · (1 - η · v)^T.