Documentation

Atlas.HighDimensionalStatistics.code.Chapter1.Lemma_1_3

theorem SubGaussian.lemma_1_3_upper_tail {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} [MeasureTheory.IsProbabilityMeasure μ] {X : Ω} {σsq : } (hX : IsSubGaussian X σsq μ) (t : ) (ht : 0 < t) :
μ {ω : Ω | X ω > t} ENNReal.ofReal (Real.exp (-(t ^ 2 / (2 * σsq))))

Upper tail bound for sub-Gaussian random variables (Lemma 1.3). For X ~ subG(σ²) and t > 0, P(X > t) ≤ exp(-t²/(2σ²)). The proof uses the Chernoff bound with the optimal choice s = t/σ².

theorem SubGaussian.lemma_1_3_lower_tail {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} [MeasureTheory.IsProbabilityMeasure μ] {X : Ω} {σsq : } (hX : IsSubGaussian X σsq μ) (t : ) (ht : 0 < t) :
μ {ω : Ω | X ω < -t} ENNReal.ofReal (Real.exp (-(t ^ 2 / (2 * σsq))))

Lower tail bound for sub-Gaussian random variables (Lemma 1.3). For X ~ subG(σ²) and t > 0, P(X < -t) ≤ exp(-t²/(2σ²)). The proof uses the lower Chernoff bound with s = -(t/σ²).

theorem SubGaussian.subGaussian_tail_bounds {Ω : Type u_1} [MeasurableSpace Ω] {μ : MeasureTheory.Measure Ω} [MeasureTheory.IsProbabilityMeasure μ] {X : Ω} {σsq : } (hX : IsSubGaussian X σsq μ) (t : ) (ht : 0 < t) :
μ {ω : Ω | X ω > t} ENNReal.ofReal (Real.exp (-(t ^ 2 / (2 * σsq)))) μ {ω : Ω | X ω < -t} ENNReal.ofReal (Real.exp (-(t ^ 2 / (2 * σsq))))

Lemma 1.3: Sub-Gaussian tail bounds. For X ~ subG(σ²) and t > 0, both P(X > t) ≤ exp(-t²/(2σ²)) and P(X < -t) ≤ exp(-t²/(2σ²)) hold.