Professional Documents
Culture Documents
Useful Inequalities: v0.36 August 10, 2021
Useful Inequalities: v0.36 August 10, 2021
√1
Useful Inequalities {x2 > 0} v0.36 · August 10, 2021 square root 2 x+1−2 x < x
< x+1− x−1 < 2 x−2 x−1 for x ≥ 1.
x x2 √ x
n
2 n
n
1− 2
− 2
≤ 1−x≤1− 2
for x ≤ 1.
x2i yi2
P P P
Cauchy-Schwarz xi y i ≤
k (n−k+1)k n nk en k n nn
i=1 i=1 i=1 max { n
binomial kk
, k!
} ≤ k
≤ k!
≤ k
; k
≤ kk (n−k)n−k
.
n
1 n
1 n
1 √
nk n n
≤ n for n ≥ k ≥ 0; √4πn (1 − 8n 1
) ≤ 2n ≤ √4πn (1 − 9n
1
p p p
Minkowski
P
|xi + yi |p ≤
P
|xi |p +
P
|yi |p for p ≥ 1. 4k! k n
).
i=1 i=1 i=1
n1 n2
≤ nk1 +k
+n2 tn
≥ tk n
k1 k2
; k k
for t ≥ 1.
1/p 1/q 1 2
n n n
√
1 1 nH(α)
|xi |p |yi |q π
P P P
|xi yi | ≤ for p, q > 1, + = 1. n
Hölder ≤ G for G = √ 2 , H(x) = − log2 (xx (1−x)1−x ).
p q 2
G ≤ αn
i=1 i=1 i=1 2πnα(1−α)
n o
Pd n d en d
, 2n
Bernoulli (1 + x)r ≥ 1 + rx for x ≥ −1, r ∈ R \ (0, 1). Reverse for r ∈ [0, 1]. i=0 i ≤ min n + 1, d
for n ≥ d ≥ 1.
(1 + x)r ≤1+ (2r − 1)x for x ∈ [0, 1], r ∈ R \ (0, 1).
2 o
1
Pαn n n
1−α n nH(α) , 2n e−2n 2 −α
i=0 i ≤ min 1−2α αn , 2 for α ∈ (0, 12 ).
1
(1 + x)n ≤ 1−nx
for x ∈ [−1, 0], n ∈ N.
n n
√ n n 1/(12n+1)
√ n n 1/12n
n n
rx 1 e ≤ 2πn e ≤ n! ≤ 2πn e ≤ en
(1 + x)r ≤ 1 + 1−(r−1)x
for x ∈ [−1, r−1
), r > 1. Stirling e e e e
Pn Pn
Karamata i=1 ϕ(ai ) ≥ i=1 ϕ(bi )
for a1 ≥ a2 ≥ · · · ≥ an , b1 ≥ · · · ≥ bn ,
f (b)−f (a)
Cauchy f ′ (a) ≤ b−a
≤ f ′ (b) where a < b, and f convex. and {ai } {bi } (majorization), i.e. ti=1 ai ≥ ti=1 bi for all 1 ≤ t ≤ n,
P P
Pn Pn
with i=1 ai = i=1 bi , and ϕ convex (for concave ϕ the reverse holds).
a+b 1
Rb ϕ(a)+ϕ(b)
Hermite ϕ 2
≤ b−a a ϕ(x) dx ≤ 2
for ϕ convex.
xa 1
· · · xa xbπ(1)
1
· · · xbπ(n)
P n
P n
Muirhead π π(1) π(n)
≥ π , sums over permut. π of [n],
P ai a
Gibbs i ai log bi
≥ a log b
for ai , bi ≥ 0, or more generally: where a1 ≥ · · · ≥ an , b 1 ≥ · · · ≥ bn , {ak } {bk }, xi ≥ 0.
P bi b
P P
ai ϕ ≤aϕ for ϕ concave, and a = ai , b = bi . P∞ P∞ am bn P∞ 1 P∞ 1
i ai a Hilbert m=1 n=1 m+n ≤π m=1 a2m 2 2 2
n=1 bn for am , bn ∈ R.
n n n With max{m, n} instead of m + n, we have 4 instead of π.
ai
ai a i ≥ ai aπ(i)
P Q Q
Chong aπ(i)
≥n and for ai > 0.
i=1 i=1 i=1 P∞ a1 +a2 +···+an p p p P∞ p
Hardy n=1 n
≤ p−1 n=1 an for an ≥ 0, p > 1.
Schur-Vornicu f (x)(x − y)k (x − z)k + f (y)(y − z)k (y − x)k + f (z)(z − x)k (z − y)k ≥ 0
1 P∞ 2n 1
where x, y, z ≥ 0, k ≥ 1 integer, f convex or monotonic, f ≥ 0. Mathieu c2 +1/2
< n=1 (n2 +c2 )2 < c2
for c 6= 0.
p q
Young ( px1p + qy1q )−1 ≤ xy ≤ xp + yq for x, y, p, q > 0, p1 + 1q = 1.
2−c(i) ≤ 1
P
Ra R b −1 Kraft for c(i) depth of leaf i of binary tree, sum over all leaves.
0 f (x) dx + 0 f (x) dx ≥ ab, for f cont., strictly increasing.
n −1
≤ 1, A ⊂ 2[n] , no set in A is subset of another set in A.
P
n
P xi n
LYM |X|
Shapiro xi+1 +xi+2
≥ 2
where xi > 0, (xn+1 , xn+2 ) := (x1 , x2 ), X∈A
i=1
and n ≤ 12 if even, n ≤ 23 if odd. FKG Pr[x ∈ A ∩ B] ≥ Pr[x ∈ A] · Pr[x ∈ B], for A, B monotone set systems.
n n
(det A)2 ≤ A2ij
Q P
Hadamard where A is an n × n matrix. Shearer |A|t ≤
Q
|traceF (A)| for A, F ⊆ 2[n] , where every i ∈ [n]
i=1 j=1 F ∈F
1
E[Xi ] = 0, σ 2 = n
P
3/2 Var[Xi ], |Xi | ≤ M (w. prob. 1), ε ≥ 0,
E X2
4th moment where 0 < E X 4 < ∞.
E |X| ≥ 1/2 θ(u) = (1 + u) log(1 + u) − u.
E [X 4 ]
n −ε2
P
Chernoff Pr[X ≥ t] ≤ F (a)/at for X r.v., Pr[X = k] = pk , Bernstein Pr Xi ≥ ε ≤ exp 2
for Xi i.r.v.,
i=1 2(nσ + M ε/3)
F (z) = k pk z k probability gen. func., and a ≥ 1.
P
1 P
E[Xi ] = 0, |Xi | < M (w. prob. 1) for all i, σ 2 = n Var[Xi ], ε ≥ 0.
µ
eδ −µδ 2
−δ 2
Pr X ≥ (1 + δ)µ ≤ ≤ exp
(1 + δ)(1+δ) 3 Azuma Pr Xn − X0 ≥ δ ≤ 2 exp for martingale (Xk ) s.t.
2 n 2
P
P i=1 ci
for Xi i.r.v. from [0,1], X = Xi , µ = E[X], δ ≥ 0 resp. δ ∈ [0, 1).
Xi − Xi−1 < ci (w. prob. 1), for i = 1, . . . , n, δ ≥ 0.
µ
e−δ −µδ 2
Pr X ≤ (1 − δ)µ ≤ ≤ exp for δ ∈ [0, 1). n
(1 − δ)(1−δ)
2
2 Efron-Stein Var[Z] ≤ 1
E
P
Z − Z (i) for Xi , Xi ′ ∈ X i.r.v.,
2
Further from the mean: Pr X ≥ R ≤ 2−R for R ≥ 2eµ (≈ 5.44µ). i=1
n
f : X n → R, Z = f (X1 , . . . , Xn ), Z (i) = f (X1 , . . . , Xi ′ , . . . , Xn ).
k
pk P
Pr X ≥ t ≤ for Xi ∈ {0, 1} k-wise i.r.v., E[Xi ] = p, X = Xi .
−2δ 2
t
for Xi , Xi ′ ∈ X i.r.v.,
k McDiarmid Pr Z − E[Z] ≥ δ ≤ 2 exp Pn 2
i=1 ci
n k̂ (1+δ)µ
Pr X ≥ (1 + δ)µ ≤ p / for Xi ∈ [0, 1] k-wise i.r.v., Z, Z (i) as before, s.t. Z − Z (i) ≤ ci for all i, and δ ≥ 0.
k̂ k̂
P µ
k ≥ k̂ = ⌈µδ/(1 − p)⌉, E[Xi ] = pi , X = Xi , µ = E[X], p = n , δ > 0. V
∆
Janson M ≤ Pr B i ≤ M exp where Pr[Bi ] ≤ ε for all i,
−2δ 2
2 − 2ε
Hoeffding Pr X − E[X] ≥ δ ≤ 2 exp Pn 2
for Xi i.r.v., Q
M = (1 − Pr[Bi ]), ∆ =
P
Pr[Bi ∧ Bj ].
i=1 (bi − ai ) i6=j,Bi ∼Bj
P
Xi ∈ [ai , bi ] (w. prob. 1), X = Xi , δ ≥ 0. V Q Q
Lovász Pr B i ≥ (1 − xi ) > 0 where Pr[Bi ] ≤ xi · (1 − xj ),
A related lemma, assuming E[X] = 0, X ∈ [a, b] (w. prob. 1) and λ ∈ R: (i,j)∈D
2
λ (b − a)2
E eλX ≤ exp
for xi ∈ [0, 1) for all i = 1, . . . , n and D the dependency graph.
8 If each Bi mutually indep. of all other events, except at most d,
Pr max |Sk | ≥ ε ≤ ε12 Var[Sn ] = ε12
V
Pr[Bi ] ≤ p for all i = 1, . . . , n , then if ep(d + 1) ≤ 1 then Pr B i > 0.
P
Kolmogorov Var[Xi ]
k i
where X1 , . . . , Xn are i.r.v., E[Xi ] = 0,
Var[Xi ] < ∞ for all i, Sk = ki=1 Xi and ε > 0.
P
c b a László Kozma · latest version: http://www.Lkozma.net/inequalities_cheat_sheet