`Bandits.UCB.prob_ucbIndex_le`🔗

This page has the declaration's own card below, then its dependency graph, then a card for each dependency (type dependencies first, then the rest of the transitive closure). For a theorem, the graph and the dependency cards only follow its statement's dependencies (its proof is replaced by sorry, so what it proves doesn't depend on how); for everything else, both the type and the body/value are followed, since their content is part of what later declarations build on.

Minimal Lean file

`prob_ucbIndex_le`🔗

LemmaBandits.UCB.prob_ucbIndex_le

Details

No docstring.

theorem

Bandits.UCB.prob_ucbIndex_le.{u_1} {K : ℕ} {c : ℝ}
  {ν : ProbabilityTheory.Kernel (Fin K) ℝ}
  [ProbabilityTheory.IsMarkovKernel ν] {Ω : Type u_1}
  {mΩ : MeasurableSpace Ω} {P : MeasureTheory.Measure Ω}
  [MeasureTheory.IsProbabilityMeasure P] {A : ℕ → Ω → Fin K}
  {R : ℕ → Ω → ℝ} {σ2 : NNReal} [Nonempty (Fin K)]
  {alg : Learning.Algorithm (Fin K) ℝ}
  (h : Learning.IsAlgEnvSeq A R alg (Learning.stationaryEnv ν) P)
  (hν :
    ∀ (a : Fin K),
      ProbabilityTheory.HasSubgaussianMGF
        (fun x => x - ∫ (x : ℝ), id x ∂ν a) σ2 (ν a))
  (hσ2 : σ2 ≠ 0) (hc : 0 ≤ c) (a : Fin K) (n : ℕ) :
  P
      {h |
        0 < Learning.pullCount A a n h ∧
          Learning.empMean A R a n h + ucbWidth A (c * ↑σ2) a n h ≤
            ∫ (x : ℝ), id x ∂ν a} ≤
    1 / (↑n + 1) ^ (c - 1)
Bandits.UCB.prob_ucbIndex_le.{u_1} {K : ℕ}
  {c : ℝ}
  {ν : ProbabilityTheory.Kernel (Fin K) ℝ}
  [ProbabilityTheory.IsMarkovKernel ν]
  {Ω : Type u_1} {mΩ : MeasurableSpace Ω}
  {P : MeasureTheory.Measure Ω}
  [MeasureTheory.IsProbabilityMeasure P]
  {A : ℕ → Ω → Fin K} {R : ℕ → Ω → ℝ}
  {σ2 : NNReal} [Nonempty (Fin K)]
  {alg : Learning.Algorithm (Fin K) ℝ}
  (h :
    Learning.IsAlgEnvSeq A R alg
      (Learning.stationaryEnv ν) P)
  (hν :
    ∀ (a : Fin K),
      ProbabilityTheory.HasSubgaussianMGF
        (fun x =>
          x - ∫ (x : ℝ), id x ∂ν a)
        σ2 (ν a))
  (hσ2 : σ2 ≠ 0) (hc : 0 ≤ c) (a : Fin K)
  (n : ℕ) :
  P
      {h |
        0 < Learning.pullCount A a n h ∧
          Learning.empMean A R a n h +
              ucbWidth A (c * ↑σ2) a n h ≤
            ∫ (x : ℝ), id x ∂ν a} ≤
    1 / (↑n + 1) ^ (c - 1)

Code

lemma prob_ucbIndex_le [Nonempty (Fin K)] {alg : Algorithm (Fin K) ℝ}
    (h : IsAlgEnvSeq A R alg (stationaryEnv ν) P)
    (hν : ∀ a, HasSubgaussianMGF (fun x ↦ x - (ν a)[id]) σ2 (ν a))
    (hσ2 : σ2 ≠ 0) (hc : 0 ≤ c) (a : Fin K) (n : ℕ) :
    P {h | 0 < pullCount A a n h ∧ empMean A R a n h + ucbWidth A (c * σ2) a n h ≤ (ν a)[id]} ≤
      1 / (n + 1) ^ (c - 1)

Type uses (6)

Body uses (3)

Used by (2)

Actions: Source · Open Issue

Proof

by
  let s : Set (ℕ × ℝ) := {(m, x) | 0 < m ∧ x / m + √(2 * (c * σ2) * log (↑n + 1) / m) ≤ (ν a)[id]}
  have hs : MeasurableSet s := by
    simp only [Nat.cast_nonneg, sqrt_div', id_eq, measurableSet_setOf, s]
    fun_prop
  classical
  calc P {h | 0 < pullCount A a n h ∧ empMean A R a n h + ucbWidth A (c * σ2) a n h ≤ (ν a)[id]}
  _ ≤ ∑ k ∈ range (n + 1) with k ∈ Prod.fst '' s,
      (streamMeasure ν) {ω | ∑ i ∈ range k, ω i a ∈ Prod.mk k ⁻¹' s} :=
    prob_pullCount_prod_sumRewards_mem_le h hs
  _ ≤ ∑ k ∈ Icc 1 n,
      (streamMeasure ν) {ω | ∑ i ∈ range k, ω i a ∈ Prod.mk k ⁻¹' s} := by
    refine Finset.sum_le_sum_of_subset_of_nonneg (fun m ↦ ?_) fun _ _ _ ↦ by positivity
    simp [s]
    grind
  _ = ∑ k ∈ Icc 1 n,
      (streamMeasure ν) {ω | (∑ i ∈ range k, ω i a) / k + √(2 * c * σ2 * log (↑n + 1) / k) ≤
        (ν a)[id]} := by
    refine Finset.sum_congr rfl fun k hk ↦ ?_
    congr with ω
    have hk : 0 < k := by grind
    simp only [Nat.cast_nonneg, sqrt_div', id_eq, Set.preimage_setOf_eq, hk, true_and,
      Set.mem_setOf_eq, s]
    grind
  _ ≤ ∑ k ∈ Icc 1 n, (1 : ℝ≥0∞) / (n + 1) ^ c := by
    gcongr with k hk
    exact prob_avg_add_sqrt_log_le hν hσ2 hc a n k (by grind)
  _ ≤ (n + 1) * (1 : ℝ≥0∞) / (n + 1) ^ c := by
    simp only [one_div, sum_const, Nat.card_Icc, add_tsub_cancel_right, nsmul_eq_mul, mul_one]
    rw [div_eq_mul_inv ((n : ℝ≥0∞) + 1)]
    gcongr
    exact le_self_add
  _ = 1 / (n + 1) ^ (c - 1) := by
    simp only [mul_one, one_div]
    rw [ENNReal.rpow_sub _ _ (by simp) (by finiteness), ENNReal.rpow_one, div_eq_mul_inv,
      ENNReal.div_eq_inv_mul, ENNReal.mul_inv (by simp) (by simp), inv_inv]

Dependency graph

Type dependencies (6)

`Algorithm`🔗

StructureLearning.Algorithm

Details

A stochastic, sequential algorithm.

structure

Learning.Algorithm.{u_4, u_5} (𝓐 : Type u_4) (𝓨 : Type u_5)
  [MeasurableSpace 𝓐] [MeasurableSpace 𝓨] : Type (max u_4 u_5)
Learning.Algorithm.{u_4, u_5}
  (𝓐 : Type u_4) (𝓨 : Type u_5)
  [MeasurableSpace 𝓐]
  [MeasurableSpace 𝓨] : Type (max u_4 u_5)

Code

structure Algorithm (𝓐 𝓨 : Type*) [MeasurableSpace 𝓐] [MeasurableSpace 𝓨] where
  /-- Policy or sampling rule: distribution of the next action. -/
  policy : (n : ℕ) → Kernel (Iic n → 𝓐 × 𝓨) 𝓐
  /-- The policy is a Markov kernel. -/
  [h_policy : ∀ n, IsMarkovKernel (policy n)]
  /-- Distribution of the first action. -/
  p0 : Measure 𝓐
  /-- The first action distribution is a probability measure. -/
  [hp0 : IsProbabilityMeasure p0]

Used by (216)

Actions: Source · Open Issue

`IsAlgEnvSeq`🔗

StructureLearning.IsAlgEnvSeq

Details

An algorithm-environment sequence: a sequence of actions and feedbacks generated by an algorithm interacting with an environment.

structure

Learning.IsAlgEnvSeq.{u_1, u_2, u_3} {𝓐 : Type u_1} {𝓨 : Type u_2}
  {Ω : Type u_3} {m𝓐 : MeasurableSpace 𝓐} {m𝓨 : MeasurableSpace 𝓨}
  {mΩ : MeasurableSpace Ω} (A : ℕ → Ω → 𝓐) (Y : ℕ → Ω → 𝓨)
  (alg : Algorithm 𝓐 𝓨) (env : Environment 𝓐 𝓨)
  (P : MeasureTheory.Measure Ω) [MeasureTheory.IsFiniteMeasure P] : Prop
Learning.IsAlgEnvSeq.{u_1, u_2, u_3}
  {𝓐 : Type u_1} {𝓨 : Type u_2}
  {Ω : Type u_3} {m𝓐 : MeasurableSpace 𝓐}
  {m𝓨 : MeasurableSpace 𝓨}
  {mΩ : MeasurableSpace Ω} (A : ℕ → Ω → 𝓐)
  (Y : ℕ → Ω → 𝓨) (alg : Algorithm 𝓐 𝓨)
  (env : Environment 𝓐 𝓨)
  (P : MeasureTheory.Measure Ω)
  [MeasureTheory.IsFiniteMeasure P] : Prop

Code

structure IsAlgEnvSeq
    (A : ℕ → Ω → 𝓐) (Y : ℕ → Ω → 𝓨) (alg : Algorithm 𝓐 𝓨) (env : Environment 𝓐 𝓨)
    (P : Measure Ω) [IsFiniteMeasure P] : Prop where
  /-- The action sequence is measurable. -/
  measurable_action n : Measurable (A n) := by fun_prop
  /-- The feedback sequence is measurable. -/
  measurable_feedback n : Measurable (Y n) := by fun_prop
  /-- The first action has the correct law. -/
  hasLaw_action_zero : HasLaw (fun ω ↦ (A 0 ω)) alg.p0 P
  /-- The first feedback has the correct conditional distribution. -/
  hasCondDistrib_feedback_zero : HasCondDistrib (Y 0) (A 0) env.ν0 P
  /-- The next action has the correct conditional distribution given the history. -/
  hasCondDistrib_action n :
    HasCondDistrib (A (n + 1)) (history A Y n) (alg.policy n) P
  /-- The next feedback has the correct conditional distribution given the history and
  next action. -/
  hasCondDistrib_feedback n :
    HasCondDistrib (Y (n + 1)) (fun ω ↦ (history A Y n ω, A (n + 1) ω))
      (env.feedback n) P

Type uses (3)

Used by (111)

Actions: Source · Open Issue

`stationaryEnv`🔗

DefinitionLearning.stationaryEnv

Details

A stationary environment, in which the distribution of the next feedback depends only on the last action.

def

Learning.stationaryEnv.{u_1, u_2} {𝓐 : Type u_1} {𝓨 : Type u_2}
  {m𝓐 : MeasurableSpace 𝓐} {m𝓨 : MeasurableSpace 𝓨}
  (ν : ProbabilityTheory.Kernel 𝓐 𝓨)
  [ProbabilityTheory.IsMarkovKernel ν] : Environment 𝓐 𝓨
Learning.stationaryEnv.{u_1, u_2}
  {𝓐 : Type u_1} {𝓨 : Type u_2}
  {m𝓐 : MeasurableSpace 𝓐}
  {m𝓨 : MeasurableSpace 𝓨}
  (ν : ProbabilityTheory.Kernel 𝓐 𝓨)
  [ProbabilityTheory.IsMarkovKernel ν] :
  Environment 𝓐 𝓨

Code

def stationaryEnv (ν : Kernel 𝓐 𝓨) [IsMarkovKernel ν] : Environment 𝓐 𝓨 := obliviousEnv fun _ ↦ ν

Type uses (1)

Environment

Body uses (1)

obliviousEnv

Used by (81)

Actions: Source · Open Issue

`pullCount`🔗

DefinitionLearning.pullCount

Details

Number of times action a was chosen up to time t (excluding t).

def

Learning.pullCount.{u_1, u_3} {𝓐 : Type u_1} {Ω : Type u_3}
  [DecidableEq 𝓐] (A : ℕ → Ω → 𝓐) (a : 𝓐) (t : ℕ) (ω : Ω) : ℕ
Learning.pullCount.{u_1, u_3}
  {𝓐 : Type u_1} {Ω : Type u_3}
  [DecidableEq 𝓐] (A : ℕ → Ω → 𝓐) (a : 𝓐)
  (t : ℕ) (ω : Ω) : ℕ

Code

noncomputable
def pullCount (A : ℕ → Ω → 𝓐) (a : 𝓐) (t : ℕ) (ω : Ω) : ℕ :=
  #(filter (fun s ↦ A s ω = a) (range t))

Used by (146)

Actions: Source · Open Issue

`empMean`🔗

DefinitionLearning.empMean

Details

Empirical mean reward obtained when pulling action a up to time t (exclusive).

def

Learning.empMean.{u_1, u_3} {𝓐 : Type u_1} {Ω : Type u_3}
  [DecidableEq 𝓐] (A : ℕ → Ω → 𝓐) (R' : ℕ → Ω → ℝ) (a : 𝓐) (t : ℕ)
  (ω : Ω) : ℝ
Learning.empMean.{u_1, u_3} {𝓐 : Type u_1}
  {Ω : Type u_3} [DecidableEq 𝓐]
  (A : ℕ → Ω → 𝓐) (R' : ℕ → Ω → ℝ) (a : 𝓐)
  (t : ℕ) (ω : Ω) : ℝ

Code

noncomputable
def empMean (A : ℕ → Ω → 𝓐) (R' : ℕ → Ω → ℝ) (a : 𝓐) (t : ℕ) (ω : Ω) : ℝ :=
  sumRewards A R' a t ω / pullCount A a t ω

Body uses (2)

Used by (34)

Actions: Source · Open Issue

`ucbWidth`🔗

DefinitionBandits.UCB.ucbWidth

Details

The exploration bonus of the UCB algorithm, which corresponds to the width of a confidence interval.

def

Bandits.UCB.ucbWidth.{u_1} {K : ℕ} {Ω : Type u_1} (A : ℕ → Ω → Fin K)
  (c : ℝ) (a : Fin K) (n : ℕ) (ω : Ω) : ℝ
Bandits.UCB.ucbWidth.{u_1} {K : ℕ}
  {Ω : Type u_1} (A : ℕ → Ω → Fin K)
  (c : ℝ) (a : Fin K) (n : ℕ) (ω : Ω) : ℝ

Code

noncomputable def ucbWidth (A : ℕ → Ω → Fin K) (c : ℝ) (a : Fin K) (n : ℕ) (ω : Ω) : ℝ :=
  √(2 * c * log (n + 1) / pullCount A a n ω)

Body uses (1)

pullCount

Used by (16)

Actions: Source · Open Issue

All dependencies, transitively (4)

`Environment`🔗

StructureLearning.Environment

Details

A stochastic environment.

structure

Learning.Environment.{u_4, u_5} (𝓐 : Type u_4) (𝓨 : Type u_5)
  [MeasurableSpace 𝓐] [MeasurableSpace 𝓨] : Type (max u_4 u_5)
Learning.Environment.{u_4, u_5}
  (𝓐 : Type u_4) (𝓨 : Type u_5)
  [MeasurableSpace 𝓐]
  [MeasurableSpace 𝓨] : Type (max u_4 u_5)

Code

structure Environment (𝓐 𝓨 : Type*) [MeasurableSpace 𝓐] [MeasurableSpace 𝓨] where
  /-- Distribution of the next observation as function of the past history. -/
  feedback : (n : ℕ) → Kernel ((Iic n → 𝓐 × 𝓨) × 𝓐) 𝓨
  /-- The feedback kernels are Markov kernels. -/
  [h_feedback : ∀ n, IsMarkovKernel (feedback n)]
  /-- Distribution of the first observation given the first action. -/
  ν0 : Kernel 𝓐 𝓨
  /-- The initial observation kernel is a Markov kernel. -/
  [hp0 : IsMarkovKernel ν0]

Used by (128)

Actions: Source · Open Issue

`history`🔗

DefinitionLearning.history

Details

History of the algorithm-environment sequence up to time n.

def

Learning.history.{u_1, u_2, u_3} {𝓐 : Type u_1} {𝓨 : Type u_2}
  {Ω : Type u_3} (A : ℕ → Ω → 𝓐) (Y : ℕ → Ω → 𝓨) (n : ℕ) (ω : Ω) :
  ↥(Finset.Iic n) → 𝓐 × 𝓨
Learning.history.{u_1, u_2, u_3}
  {𝓐 : Type u_1} {𝓨 : Type u_2}
  {Ω : Type u_3} (A : ℕ → Ω → 𝓐)
  (Y : ℕ → Ω → 𝓨) (n : ℕ) (ω : Ω) :
  ↥(Finset.Iic n) → 𝓐 × 𝓨

Code

def history (A : ℕ → Ω → 𝓐) (Y : ℕ → Ω → 𝓨) (n : ℕ) (ω : Ω) : Iic n → 𝓐 × 𝓨 :=
  fun i ↦ (A i ω, Y i ω)

Used by (72)

Actions: Source · Open Issue

`obliviousEnv`🔗

DefinitionLearning.obliviousEnv

Details

An oblivious environment, in which the distribution of the next feedback depends only on the last action, but in a possibly time-dependent manner.

def

Learning.obliviousEnv.{u_1, u_2} {𝓐 : Type u_1} {𝓨 : Type u_2}
  {m𝓐 : MeasurableSpace 𝓐} {m𝓨 : MeasurableSpace 𝓨}
  (ν : ℕ → ProbabilityTheory.Kernel 𝓐 𝓨)
  [∀ (n : ℕ), ProbabilityTheory.IsMarkovKernel (ν n)] : Environment 𝓐 𝓨
Learning.obliviousEnv.{u_1, u_2}
  {𝓐 : Type u_1} {𝓨 : Type u_2}
  {m𝓐 : MeasurableSpace 𝓐}
  {m𝓨 : MeasurableSpace 𝓨}
  (ν : ℕ → ProbabilityTheory.Kernel 𝓐 𝓨)
  [∀ (n : ℕ),
      ProbabilityTheory.IsMarkovKernel
        (ν n)] :
  Environment 𝓐 𝓨

Code

def obliviousEnv (ν : ℕ → Kernel 𝓐 𝓨) [∀ n, IsMarkovKernel (ν n)] : Environment 𝓐 𝓨 where
  feedback n := (ν (n + 1)).prodMkLeft _
  ν0 := ν 0

Type uses (1)

Environment

Used by (10)

Actions: Source · Open Issue

`sumRewards`🔗

DefinitionLearning.sumRewards

Details

Sum of rewards obtained when pulling action a up to time t (exclusive).

def

Learning.sumRewards.{u_1, u_3} {𝓐 : Type u_1} {Ω : Type u_3}
  [DecidableEq 𝓐] (A : ℕ → Ω → 𝓐) (R' : ℕ → Ω → ℝ) (a : 𝓐) (t : ℕ)
  (ω : Ω) : ℝ
Learning.sumRewards.{u_1, u_3}
  {𝓐 : Type u_1} {Ω : Type u_3}
  [DecidableEq 𝓐] (A : ℕ → Ω → 𝓐)
  (R' : ℕ → Ω → ℝ) (a : 𝓐) (t : ℕ)
  (ω : Ω) : ℝ

Code

def sumRewards (A : ℕ → Ω → 𝓐) (R' : ℕ → Ω → ℝ) (a : 𝓐) (t : ℕ) (ω : Ω) : ℝ :=
  ∑ s ∈ range t, if A s ω = a then R' s ω else 0

Used by (44)

Actions: Source · Open Issue

Bandits.UCB.prob_ucbIndex_le🔗

prob_ucbIndex_le🔗

Algorithm🔗

IsAlgEnvSeq🔗

stationaryEnv🔗

pullCount🔗

empMean🔗

ucbWidth🔗

Environment🔗

history🔗

obliviousEnv🔗

sumRewards🔗

`Bandits.UCB.prob_ucbIndex_le`🔗

`prob_ucbIndex_le`🔗

`Algorithm`🔗

`IsAlgEnvSeq`🔗

`stationaryEnv`🔗

`pullCount`🔗

`empMean`🔗

`ucbWidth`🔗

`Environment`🔗

`history`🔗

`obliviousEnv`🔗

`sumRewards`🔗