Concentration for Coulomb gases

Size: px
Start display at page:

Download "Concentration for Coulomb gases"

Transcription

1 1/32 and Coulomb transport inequalities Djalil Chafaï 1, Adrien Hardy 2, Mylène Maïda 2 1 Université Paris-Dauphine, 2 Université de Lille November 4, 2016 IHP Paris Groupe de travail MEGA

2 2/32 Motivation Outline Motivation Electrostatics Coulomb gas model Probability metrics and Coulomb transport inequality Concentration of measure for Coulomb gases

3 3/32 Motivation Concentration of measure P( F(Z ) EF(Z ) r) 2e 1 2 r 2 Sub-Gaussian concentration (Z Gaussian, F Lipschitz)

4 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N

5 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N Examples

6 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N Examples Standard: Z N = X X N

7 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N Examples Standard: Z N = X X N RMT: Z N = f (λ 1 (X)) + + f (λ N (X))

8 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N Examples Standard: Z N = X X N RMT: Z N = f (λ 1 (X)) + + f (λ N (X)) TSP: Z N = inf σ ΣN N i=1 X σ(i+1) X σ(i)

9 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N Examples Standard: Z N = X X N RMT: Z N = f (λ 1 (X)) + + f (λ N (X)) TSP: Z N = inf σ ΣN N i=1 X σ(i+1) X σ(i) High dimensional phenomena, combinatorial optimization

10 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N Examples Standard: Z N = X X N RMT: Z N = f (λ 1 (X)) + + f (λ N (X)) TSP: Z N = inf σ ΣN N i=1 X σ(i+1) X σ(i) High dimensional phenomena, combinatorial optimization Talagrand principle

11 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N Examples Standard: Z N = X X N RMT: Z N = f (λ 1 (X)) + + f (λ N (X)) TSP: Z N = inf σ ΣN N i=1 X σ(i+1) X σ(i) High dimensional phenomena, combinatorial optimization Talagrand principle Erdős complete convergence to deterministic object

12 3/32 Motivation Concentration of measure f (N,r) P( Z N EZ N r) 2e Sub-Gaussian concentration (Z Gaussian, F Lipschitz) Dependence with respect to N Examples Standard: Z N = X X N RMT: Z N = f (λ 1 (X)) + + f (λ N (X)) TSP: Z N = inf σ ΣN N i=1 X σ(i+1) X σ(i) High dimensional phenomena, combinatorial optimization Talagrand principle Erdős complete convergence to deterministic object Books: Steele, Ledoux, Boucheron-Lugosi-Massart

13 4/32 Motivation Gaussian exactly solvable model Ginibre G = (G jk ) 1 j,k N iid C Gaussian of variance 1 2N

14 4/32 Motivation Gaussian exactly solvable model Ginibre G = (G jk ) 1 j,k N iid C Gaussian of variance 1 2N The matrix G has density on C N2 e N N j,k=1 G jk 2 = e NTr(GG )

15 4/32 Motivation Gaussian exactly solvable model Ginibre G = (G jk ) 1 j,k N iid C Gaussian of variance 1 2N The matrix G has density on C N2 e N N j,k=1 G jk 2 = e NTr(GG ) Change of variable: G = UTU (U, T = D + N)

16 4/32 Motivation Gaussian exactly solvable model Ginibre G = (G jk ) 1 j,k N iid C Gaussian of variance 1 2N The matrix G has density on C N2 e N N j,k=1 G jk 2 = e NTr(GG ) Change of variable: G = UTU (U, T = D + N) Tr(GG ) = Tr(TT ) = Tr(DD ) + Tr(NN )

17 4/32 Motivation Gaussian exactly solvable model Ginibre G = (G jk ) 1 j,k N iid C Gaussian of variance 1 2N The matrix G has density on C N2 e N N j,k=1 G jk 2 = e NTr(GG ) Change of variable: G = UTU (U, T = D + N) Tr(GG ) = Tr(TT ) = Tr(DD ) + Tr(NN ) (λ 1 (G),..., λ N (G)) has density ( ) N ϕ N (z 1,..., z N ) exp N z r 2 z j z k 2. r=1 1 j<k N

18 4/32 Motivation Gaussian exactly solvable model Ginibre G = (G jk ) 1 j,k N iid C Gaussian of variance 1 2N The matrix G has density on C N2 e N N j,k=1 G jk 2 = e NTr(GG ) Change of variable: G = UTU (U, T = D + N) Tr(GG ) = Tr(TT ) = Tr(DD ) + Tr(NN ) (λ 1 (G),..., λ N (G)) has density ( ) N ϕ N (z 1,..., z N ) exp N z r 2 z j z k 2. r=1 Neither product nor log-concave 1 j<k N

19 4/32 Motivation Gaussian exactly solvable model Ginibre G = (G jk ) 1 j,k N iid C Gaussian of variance 1 2N The matrix G has density on C N2 e N N j,k=1 G jk 2 = e NTr(GG ) Change of variable: G = UTU (U, T = D + N) Tr(GG ) = Tr(TT ) = Tr(DD ) + Tr(NN ) (λ 1 (G),..., λ N (G)) has density ( ) N ϕ N (z 1,..., z N ) exp N z r 2 z j z k 2. r=1 1 j<k N Neither product nor log-concave Determinantal (Pemantle-Peres, Breuer-Duits)

20 5/32 Motivation First order global asymptotics Empirical spectral distribution µ G := 1 N N k=1 δ λ k (G)

21 5/32 Motivation First order global asymptotics Empirical spectral distribution µ G := 1 N N k=1 δ λ k (G) Mehta: density of mean empirical spectral distribution Eµ G : ϕ (1) N 1 e N z 2 N l z 2l N (z) = π l! l=0

22 5/32 Motivation First order global asymptotics Empirical spectral distribution µ G := 1 N N k=1 δ λ k (G) Mehta: density of mean empirical spectral distribution Eµ G : Mehta: mean circular law: ϕ (1) N 1 e N z 2 N l z 2l N (z) = π l! l=0 ϕ (1) N (z) N 1 [0,1] ( z ). π

23 5/32 Motivation First order global asymptotics Empirical spectral distribution µ G := 1 N N k=1 δ λ k (G) Mehta: density of mean empirical spectral distribution Eµ G : Mehta: mean circular law: ϕ (1) N 1 e N z 2 N l z 2l N (z) = π l! l=0 ϕ (1) N (z) N 1 [0,1] ( z ). π Silverstein: N E((µ Gf Eµ G f ) 4 ) < strong law: a.s. µ G w n µ.

24 5/32 Motivation First order global asymptotics Empirical spectral distribution µ G := 1 N N k=1 δ λ k (G) Mehta: density of mean empirical spectral distribution Eµ G : Mehta: mean circular law: ϕ (1) N 1 e N z 2 N l z 2l N (z) = π l! l=0 ϕ (1) N (z) N 1 [0,1] ( z ). π Silverstein: N E((µ Gf Eµ G f ) 4 ) < strong law: a.s. µ G w n µ. Open problem: concentration P(d(µ G, µ ) r)? e cn2 r 2

25 /32 Motivation Sub-Gaussian concentration of measure Gaussian Unitary Ensemble (GUE) H = (H jk ) 1 j,k N e NTr(H2 ) e N N k=1 λ2 k (λ j λ k ) 2 1 j<k N

26 /32 Motivation Sub-Gaussian concentration of measure Gaussian Unitary Ensemble (GUE) H = (H jk ) 1 j,k N e NTr(H2 ) e N N k=1 λ2 k (λ j λ k ) 2 1 j<k N Hoffman-Wielandt inequality for H, H Herm N N min σ Σ N N (λ k (H) λ σ(k) (H )) 2 k=1 N (H jk H jk )2. j,k=1

27 /32 Motivation Sub-Gaussian concentration of measure Gaussian Unitary Ensemble (GUE) H = (H jk ) 1 j,k N e NTr(H2 ) e N N k=1 λ2 k (λ j λ k ) 2 1 j<k N Hoffman-Wielandt inequality for H, H Herm N N NW 2 (µ H, µ H ) 2 H H 2 HS. Sub-Gaussian concentration inequality for GUE P( W 2 (µ H, Eµ H ) EW 2 (µ H, Eµ H ) r) 2e cn2 r 2.

28 /32 Motivation Sub-Gaussian concentration of measure Gaussian Unitary Ensemble (GUE) H = (H jk ) 1 j,k N e NTr(H2 ) e N N k=1 λ2 k (λ j λ k ) 2 1 j<k N Hoffman-Wielandt inequality for H, H Herm N N NW 2 (µ H, µ H ) 2 H H 2 HS. Sub-Gaussian concentration inequality for GUE P( W 2 (µ H, Eµ H ) EW 2 (µ H, Eµ H ) r) 2e cn2 r 2. Maïda-Maurel-Segala: P(W 1 (µ H, µ ) r) e cn2 r 2.

29 /32 Motivation Sub-Gaussian concentration of measure Open problem for Ginibre: P(W 1 (µ G, µ ) r) e cn2 r 2.

30 /32 Motivation Sub-Gaussian concentration of measure Open problem for Ginibre: P(W 1 (µ G, µ ) r) e cn2 r 2. Meckes and Meckes: E(W p (µ G, µ ))) by coupling

31 /32 Motivation Sub-Gaussian concentration of measure Open problem for Ginibre: P(W 1 (µ G, µ ) r) e cn2 r 2. Meckes and Meckes: E(W p (µ G, µ ))) by coupling Ortega-Cerda et al: E(W 1 (µ S, µ ))) by complex transport

32 /32 Motivation Sub-Gaussian concentration of measure Open problem for Ginibre: P(W 1 (µ G, µ ) r) e cn2 r 2. Meckes and Meckes: E(W p (µ G, µ ))) by coupling Ortega-Cerda et al: E(W 1 (µ S, µ ))) by complex transport Coulomb gas = Coulomb Boltzmann Gibbs measure e N N k=1 z k 2 1 j<k N z j z k 2

33 /32 Motivation Sub-Gaussian concentration of measure Open problem for Ginibre: P(W 1 (µ G, µ ) r) e cn2 r 2. Meckes and Meckes: E(W p (µ G, µ ))) by coupling Ortega-Cerda et al: E(W 1 (µ S, µ ))) by complex transport Coulomb gas = Coulomb Boltzmann Gibbs measure N exp N z k 2 1 log z j z k k=1 1 j k N Exchangeable but neither product nor log-concave

34 /32 Motivation Sub-Gaussian concentration of measure Open problem for Ginibre: P(W 1 (µ G, µ ) r) e cn2 r 2. Meckes and Meckes: E(W p (µ G, µ ))) by coupling Ortega-Cerda et al: E(W 1 (µ S, µ ))) by complex transport Coulomb gas = Coulomb Boltzmann Gibbs measure ( )) exp ( N 2 z 2 1 µ N (dz) + log z w µ N(dz)µ N (dw) Exchangeable but neither product nor log-concave Empirical measure µ N := 1 N N k=1 δ z k

35 /32 Motivation Sub-Gaussian concentration of measure Open problem for Ginibre: P(W 1 (µ G, µ ) r) e cn2 r 2. Meckes and Meckes: E(W p (µ G, µ ))) by coupling Ortega-Cerda et al: E(W 1 (µ S, µ ))) by complex transport Coulomb gas = Coulomb Boltzmann Gibbs measure ( )) exp ( N 2 V dµ N + g(z w)µ N (dz)µ N (dw) Exchangeable but neither product nor log-concave Empirical measure µ N := 1 N N k=1 δ z k

36 8/32 Electrostatics Outline Motivation Electrostatics Coulomb gas model Probability metrics and Coulomb transport inequality Concentration of measure for Coulomb gases

37 /32 Electrostatics Coulomb kernel in mathematical physics Coulomb kernel in R d, d 2, log 1 if d = 2, x R d x g(x) := 1 x d 2 if d 3.

38 /32 Electrostatics Coulomb kernel in mathematical physics Coulomb kernel in R d, d 2, log 1 if d = 2, x R d x g(x) := 1 x d 2 if d 3. Fundamental solution of Poisson s equation { 2π if d = 2, g = c d δ 0 where c d := (d 2) S d 1 if d 3.

39 0/32 Electrostatics Coulomb energy and metric Probability measures on R d with compact support

40 0/32 Electrostatics Coulomb energy and metric Probability measures on R d with compact support Coulomb energy: E(µ) := g(x y)µ(dx)µ(dy) R {+ }.

41 0/32 Electrostatics Coulomb energy and metric Probability measures on R d with compact support Coulomb energy: E(µ) := g(x y)µ(dx)µ(dy) R {+ }. Coulomb metric: (µ, ν) E(µ ν).

42 1/32 Electrostatics Confinement and equilibrium measure External potential V : R d R {+ } growing at infinity

43 1/32 Electrostatics Confinement and equilibrium measure External potential V : R d R {+ } growing at infinity Coulomb energy with confining potential E V (µ) = V (x)µ(dx) + E(µ) = (V (x) + (g µ)(x))µ(dx).

44 1/32 Electrostatics Confinement and equilibrium measure External potential V : R d R {+ } growing at infinity Coulomb energy with confining potential E V (µ) = V (x)µ(dx) + E(µ) = (V (x) + (g µ)(x))µ(dx). External and internal electric fields: V + g µ

45 1/32 Electrostatics Confinement and equilibrium measure External potential V : R d R {+ } growing at infinity Coulomb energy with confining potential E V (µ) = V (x)µ(dx) + E(µ) = (V (x) + (g µ)(x))µ(dx). External and internal electric fields: V + g µ Equilibrium probability measure µ V := arg inf E V

46 1/32 Electrostatics Confinement and equilibrium measure External potential V : R d R {+ } growing at infinity Coulomb energy with confining potential E V (µ) = V (x)µ(dx) + E(µ) = (V (x) + (g µ)(x))µ(dx). External and internal electric fields: V + g µ Equilibrium probability measure µ V := arg inf E V µ V is compactly supported and has density 1 2c d V

47 12/32 Electrostatics Examples of equilibrium measures d g V µ V interval c(x) arcsine 1 2 x 2 semicircle 2 2 x 2 uniform on a disc 3 d x 2 uniform on a ball 2 d radial radial in a ring

48 13/32 Coulomb gas model Outline Motivation Electrostatics Coulomb gas model Probability metrics and Coulomb transport inequality Concentration of measure for Coulomb gases

49 14/32 Coulomb gas model Coulomb gas or one component plasma Interaction energy of N Coulomb charges in R d : H N (x 1,..., x N ) := N N i=1 V (x i ) + i j g(x i x j ).

50 14/32 Coulomb gas model Coulomb gas or one component plasma Interaction energy of N Coulomb charges in R d : H N (x 1,..., x N ) := N N i=1 V (x i ) + i j g(x i x j ). Boltzmann Gibbs probability measure on (R d ) N dp N V,β (x 1,..., x N ) dx 1 dx N ( exp β ) 2 H N(x 1,..., x N )

51 14/32 Coulomb gas model Coulomb gas or one component plasma Interaction energy of N Coulomb charges in R d : H N (x 1,..., x N ) := N N i=1 V (x i ) + i j g(x i x j ). Boltzmann Gibbs probability measure on (R d ) N dp N V,β (x 1,..., x N ) dx 1 dx N ( exp β ) 2 H N(x 1,..., x N ) V must be strong enough at infinity to ensure integrability.

52 5/32 Coulomb gas model Empirical measure and equilibrium measure Random empirical measure under P N V,β : ˆµ N := 1 N N δ xi. i=1

53 5/32 Coulomb gas model Empirical measure and equilibrium measure Random empirical measure under P N V,β : ˆµ N := 1 N N δ xi. i=1 Under mild assumptions on V, with probability one, µ N N µ V.

54 5/32 Coulomb gas model Empirical measure and equilibrium measure Random empirical measure under P N V,β : ˆµ N := 1 N N δ xi. i=1 Under mild assumptions on V, with probability one, N 2 µ N N µ V. Large Deviation Principle (BAG, HP, BAZ, CGZ, S, B) ( ) log P N V,β d(µ N, µ V ) r β ( inf EV (µ) E V (µ V ) ). N 2 d(µ,µ V ) r

55 16/32 Coulomb gas model Quantitative or non asymptotic estimates The LDP gives for any r > 0 and any N N 0, ( ) e cr N2 P N V,β d(µ N, µ V ) r e Cr N2.

56 16/32 Coulomb gas model Quantitative or non asymptotic estimates The LDP gives for any r > 0 and any N N 0, ( ) e cr N2 P N V,β d(µ N, µ V ) r e Cr N2. Sub-Gaussian concentration of measure: C quadratic in r?

57 16/32 Coulomb gas model Quantitative or non asymptotic estimates The LDP gives for any r > 0 and any N N 0, ( ) e cr N2 P N V,β d(µ N, µ V ) r e Cr N2. Sub-Gaussian concentration of measure: C quadratic in r? Other distances such as W p?

58 16/32 Coulomb gas model Quantitative or non asymptotic estimates The LDP gives for any r > 0 and any N N 0, ( ) e cr N2 P N V,β d(µ N, µ V ) r e Cr N2. Sub-Gaussian concentration of measure: C quadratic in r? Other distances such as W p? Yes for one-dimensional log-gas: Maïda-Maurel-Segala

59 16/32 Coulomb gas model Quantitative or non asymptotic estimates The LDP gives for any r > 0 and any N N 0, ( ) e cr N2 P N V,β d(µ N, µ V ) r e Cr N2. Sub-Gaussian concentration of measure: C quadratic in r? Other distances such as W p? Yes for one-dimensional log-gas: Maïda-Maurel-Segala Nothing known otherwise (nothing for Ginibre ensemble!)

60 7/32 Coulomb gas model Key observation Write P N V,β with µ N: ( ) dp N V,β (x 1,..., x N ) exp β 2 N2 E V (µ N) = dx 1 dx N ZV N,β where E V (µ N) := V (x)µ N (dx) + g(x y)µ N (dx)µ N (dy). x y

61 7/32 Coulomb gas model Key observation Write P N V,β with µ N: ( ) dp N V,β (x 1,..., x N ) exp β 2 N2 E V (µ N) = dx 1 dx N ZV N,β where E V (µ N) := V (x)µ N (dx) + g(x y)µ N (dx)µ N (dy). x y Serfaty et al: rewrite E V (µ N) E V (µ V ) with L 2 norm of electric field of µ N µ V. Leads to renormalized energy.

62 7/32 Coulomb gas model Key observation Write P N V,β with µ N: ( ) dp N V,β (x 1,..., x N ) exp β 2 N2 E V (µ N) = dx 1 dx N ZV N,β where E V (µ N) := V (x)µ N (dx) + g(x y)µ N (dx)µ N (dy). x y Serfaty et al: rewrite E V (µ N) E V (µ V ) with L 2 norm of electric field of µ N µ V. Leads to renormalized energy. Alternative: compare E V (µ N) E V (µ V ) with W 1 (µ N, µ V ).

63 18/32 Probability metrics and Coulomb transport inequality Outline Motivation Electrostatics Coulomb gas model Probability metrics and Coulomb transport inequality Concentration of measure for Coulomb gases

64 19/32 Probability metrics and Coulomb transport inequality Probability metrics Coulomb divergence E V (µ) E V (µ V )

65 19/32 Probability metrics and Coulomb transport inequality Probability metrics Coulomb divergence E V (µ) E V (µ V ) Coulomb metric E(µ ν)

66 19/32 Probability metrics and Coulomb transport inequality Probability metrics Coulomb divergence E V (µ) E V (µ V ) Coulomb metric E(µ ν) Bounded-Lipschitz or Fortet Mourier distance d BL (µ, ν) := sup f Lip 1 f 1 f (x)(µ ν)(dx),

67 19/32 Probability metrics and Coulomb transport inequality Probability metrics Coulomb divergence E V (µ) E V (µ V ) Coulomb metric E(µ ν) Bounded-Lipschitz or Fortet Mourier distance d BL (µ, ν) := sup f Lip 1 f 1 f (x)(µ ν)(dx), (Monge-Kantorovich-)Wasserstein distance W p (µ, ν) := inf E( X Y p ) 1/p. (X,Y ) X µ,y ν

68 9/32 Probability metrics and Coulomb transport inequality Probability metrics Coulomb divergence E V (µ) E V (µ V ) Coulomb metric E(µ ν) Bounded-Lipschitz or Fortet Mourier distance d BL (µ, ν) := sup f Lip 1 f 1 f (x)(µ ν)(dx), (Monge-Kantorovich-)Wasserstein distance ( 1/p. W p (µ, ν) := x y p π(dx, dy)) inf π Π(µ,ν)

69 9/32 Probability metrics and Coulomb transport inequality Probability metrics Coulomb divergence E V (µ) E V (µ V ) Coulomb metric E(µ ν) Bounded-Lipschitz or Fortet Mourier distance d BL (µ, ν) := sup f Lip 1 f 1 f (x)(µ ν)(dx), Kantorovich-Rubinstein duality W 1 (µ, ν) = sup f Lip 1 f (x)(µ ν)(dx).

70 Probability metrics and Coulomb transport inequality Probability metrics Coulomb divergence E V (µ) E V (µ V ) Coulomb metric E(µ ν) 9/32 Bounded-Lipschitz or Fortet Mourier distance d BL (µ, ν) := sup f Lip 1 f 1 f (x)(µ ν)(dx), Kantorovich-Rubinstein duality d BL (µ, ν) W 1 (µ, ν) = sup f Lip 1 Topologies f (x)(µ ν)(dx).

71 0/32 Probability metrics and Coulomb transport inequality Local Coulomb transport inequality Theorem (Transport type inequality CHM 2016) D R d compact, supp(µ + ν) D, E(µ) < and E(ν) <, W 1 (µ, ν) 2 C D E(µ ν). Optimal C D is Vol(B 4Vol(D) )

72 0/32 Probability metrics and Coulomb transport inequality Local Coulomb transport inequality Theorem (Transport type inequality CHM 2016) D R d compact, supp(µ + ν) D, E(µ) < and E(ν) <, W 1 (µ, ν) 2 C D E(µ ν). Optimal C D is Vol(B 4Vol(D) ) Extends Popescu local free transport inequality to any d

73 21/32 Probability metrics and Coulomb transport inequality Coulomb transport inequality for equilibrium measures Theorem (Transport type inequality CHM 2016) We have for any probability measure µ ) d BL (µ, µ V ) 2 C BL (E V (µ) E V (µ V ). Moreover if V is superquadratic then W 1 (µ, µ V ) 2 ( C W1 EV (µ) E V (µ V ) ). Free transport inequalities for d = 2 and V = + on R c

74 21/32 Probability metrics and Coulomb transport inequality Coulomb transport inequality for equilibrium measures Theorem (Transport type inequality CHM 2016) We have for any probability measure µ ) d BL (µ, µ V ) 2 C BL (E V (µ) E V (µ V ). Moreover if V is superquadratic then W 1 (µ, µ V ) 2 ( C W1 EV (µ) E V (µ V ) ). Free transport inequalities for d = 2 and V = + on R c Extends Maïda-Maurel-Segala, Popescu

75 21/32 Probability metrics and Coulomb transport inequality Coulomb transport inequality for equilibrium measures Theorem (Transport type inequality CHM 2016) We have for any probability measure µ ) d BL (µ, µ V ) 2 C BL (E V (µ) E V (µ V ). Moreover if V is superquadratic then W 1 (µ, µ V ) 2 ( C W1 EV (µ) E V (µ V ) ). Free transport inequalities for d = 2 and V = + on R c Extends Maïda-Maurel-Segala, Popescu Growth condition is optimal for W 1

76 22/32 Concentration of measure for Coulomb gases Outline Motivation Electrostatics Coulomb gas model Probability metrics and Coulomb transport inequality Concentration of measure for Coulomb gases

77 3/32 Concentration of measure for Coulomb gases Concentration of measure for Coulomb gases Theorem (Concentration inequality CHM 2016) If V does not grows too fast then ( ) d BL (µ N, µ V ) r e aβn2 r 2. P N V,β Moreover if V superquadratic then W 1 instead of d BL. LDP shows that order in N is optimal

78 3/32 Concentration of measure for Coulomb gases Concentration of measure for Coulomb gases Theorem (Concentration inequality CHM 2016) If V does not grows too fast then ( ) d BL (µ N, µ V ) r e aβn2 r 2 +1 d=2 ( β 4 N log N)+bβN2 2/d +c(β)n. P N V,β Moreover if V superquadratic then W 1 instead of d BL. LDP shows that order in N is optimal Explicit constants a, b, c if V sub-quadratic

79 3/32 Concentration of measure for Coulomb gases Concentration of measure for Coulomb gases Theorem (Concentration inequality CHM 2016) If V does not grows too fast then ( ) d BL (µ N, µ V ) r e aβn2 r 2 +1 d=2 ( β 4 N log N)+bβN2 2/d +c(β)n. P N V,β Moreover if V superquadratic then W 1 instead of d BL. LDP shows that order in N is optimal Explicit constants a, b, c if V sub-quadratic Extends Maïda-Maurel-Segala bound to any dimension: P N V,β ( ) W 1 (µ N, µ V ) r e cn2 r 2, r { log N N if d = 2, N 1/d if d 3.

80 24/32 Concentration of measure for Coulomb gases Convergence in Wasserstein distance Corollary (Wasserstein convergence CHM 2016) If V superquadratic and β N β V log N N then under PN V,β N a.s. lim N W 1(µ N, µ V ) = 0.

81 25/32 Concentration of measure for Coulomb gases Convergence at mesoscopic scale Corollary (Mesoscopic convergence CHM 2016) If d = 2 then ( P N ( ) V,β d BL τ N s x 0 µ N, τx Ns 0 µ V CN s log N N ) e cn log N, Test functions are global, not local as in Rougerie-Serfaty

82 25/32 Concentration of measure for Coulomb gases Convergence at mesoscopic scale Corollary (Mesoscopic convergence CHM 2016) If d = 2 then ( P N ( ) V,β d BL τ N s x 0 µ N, τx Ns 0 µ V CN s log N N ) e cn log N, If d 3 then ( ( ) ) d BL τ N s x 0 µ N, τx Ns 0 µ V CN s 1/d e cn2 2/d. P N V,β Test functions are global, not local as in Rougerie-Serfaty

83 25/32 Concentration of measure for Coulomb gases Convergence at mesoscopic scale Corollary (Mesoscopic convergence CHM 2016) If d = 2 then ( P N ( ) V,β d BL τ N s x 0 µ N, τx Ns 0 µ V CN s log N N ) e cn log N, If d 3 then ( ( ) ) d BL τ N s x 0 µ N, τx Ns 0 µ V CN s 1/d e cn2 2/d. P N V,β If V superquadratic then d BL can be replaced by W 1. Test functions are global, not local as in Rougerie-Serfaty

84 6/32 Concentration of measure for Coulomb gases Concentration for spectrum of Ginibre matrices Corollary (Concentration for Ginibre CHM 2016) If G is N N with iid Gaussian entries of variance 1 2N then ( ) P W 1 (µ G, µ ) r e 1 4C N2 r N log N+N[ 1 C log π]. Open problem: universality, even for ±1

85 6/32 Concentration of measure for Coulomb gases Concentration for spectrum of Ginibre matrices Corollary (Concentration for Ginibre CHM 2016) If G is N N with iid Gaussian entries of variance 1 2N then ( ) P W 1 (µ G, µ ) r e 1 4C N2 r N log N+N[ 1 C log π]. Open problem: universality, even for ±1 Provides W 1 convergence

86 27/32 Concentration of measure for Coulomb gases Exponential tightness Theorem (Tightness CHM 2016) For any r r 0 P N V,β (supp(µ N) B r ) = P N V,β( max 1 i N x i r ) e cnv (r), where V (r) := min x r V (x). Follows by using an argument by Borot and Guionnet

87 27/32 Concentration of measure for Coulomb gases Exponential tightness Theorem (Tightness CHM 2016) For any r r 0 P N V,β (supp(µ N) B r ) = P N V,β( max 1 i N x i r ) e cnv (r), where V (r) := min x r V (x). Follows by using an argument by Borot and Guionnet Gives that almost surely lim N max 1 i N x i <.

88 27/32 Concentration of measure for Coulomb gases Exponential tightness Theorem (Tightness CHM 2016) For any r r 0 P N V,β (supp(µ N) B r ) = P N V,β( max 1 i N x i r ) e cnv (r), where V (r) := min x r V (x). Follows by using an argument by Borot and Guionnet Gives that almost surely lim N max 1 i N x i <. Gives W p versions of convergence and concentration W p p(µ, ν) (2M) p 1 W 1 (µ, ν) M(2M) p 1 d BL (µ, ν). For p = 2 we get P N V,β (W 2(µ N, µ V ) r) 2e cn3/2 r 2.

89 28/32 Concentration of measure for Coulomb gases Notes and comments W p 2 versions? Popescu free transport inequalities

90 28/32 Concentration of measure for Coulomb gases Notes and comments W p 2 versions? Popescu free transport inequalities Hardy-Littlewood-Sobolev inequalities (Keller-Segel PDE)

91 28/32 Concentration of measure for Coulomb gases Notes and comments W p 2 versions? Popescu free transport inequalities Hardy-Littlewood-Sobolev inequalities (Keller-Segel PDE) Classical transport inequalities with Coulomb distance

92 28/32 Concentration of measure for Coulomb gases Notes and comments W p 2 versions? Popescu free transport inequalities Hardy-Littlewood-Sobolev inequalities (Keller-Segel PDE) Classical transport inequalities with Coulomb distance Varying V and conditional gases. P N V,β ( x N) = P N 1 Ṽ N,β with Ṽ N := N N 1 V + 2 N 1 g(x N ) [covered by our work since g is superharmonic]

93 28/32 Concentration of measure for Coulomb gases Notes and comments W p 2 versions? Popescu free transport inequalities Hardy-Littlewood-Sobolev inequalities (Keller-Segel PDE) Classical transport inequalities with Coulomb distance Varying V and conditional gases. P N V,β ( x N) = P N 1 Ṽ N,β with Ṽ N := N N 1 V + 2 N 1 g(x N ) [covered by our work since g is superharmonic] Usage for CLT with GFF in all dimensions (VR, M+, LS, B+)

94 28/32 Concentration of measure for Coulomb gases Notes and comments W p 2 versions? Popescu free transport inequalities Hardy-Littlewood-Sobolev inequalities (Keller-Segel PDE) Classical transport inequalities with Coulomb distance Varying V and conditional gases. P N V,β ( x N) = P N 1 Ṽ N,β with Ṽ N := N N 1 V + 2 N 1 g(x N ) [covered by our work since g is superharmonic] Usage for CLT with GFF in all dimensions (VR, M+, LS, B+) Weakly confining potentials and heavy-tailed µ V

95 28/32 Concentration of measure for Coulomb gases Notes and comments W p 2 versions? Popescu free transport inequalities Hardy-Littlewood-Sobolev inequalities (Keller-Segel PDE) Classical transport inequalities with Coulomb distance Varying V and conditional gases. P N V,β ( x N) = P N 1 Ṽ N,β with Ṽ N := N N 1 V + 2 N 1 g(x N ) [covered by our work since g is superharmonic] Usage for CLT with GFF in all dimensions (VR, M+, LS, B+) Weakly confining potentials and heavy-tailed µ V Universality of concentration for random matrices

96 28/32 Concentration of measure for Coulomb gases Notes and comments W p 2 versions? Popescu free transport inequalities Hardy-Littlewood-Sobolev inequalities (Keller-Segel PDE) Classical transport inequalities with Coulomb distance Varying V and conditional gases. P N V,β ( x N) = P N 1 Ṽ N,β with Ṽ N := N N 1 V + 2 N 1 g(x N ) [covered by our work since g is superharmonic] Usage for CLT with GFF in all dimensions (VR, M+, LS, B+) Weakly confining potentials and heavy-tailed µ V Universality of concentration for random matrices Crossover and Sanov regime (Allez-Bouchaud-Guionnet)

97 29/32 Concentration of measure for Coulomb gases That s all folks! Thank you for your attention.

98 0/32 Concentration of measure for Coulomb gases Idea of proof of Coulomb transport inequality Potential: if U µ (x) := g µ(x) then U µ (x) = c d µ

99 0/32 Concentration of measure for Coulomb gases Idea of proof of Coulomb transport inequality Potential: if U µ (x) := g µ(x) then U µ (x) = c d µ Electric field: U µ (x). Carré du champ : U µ 2

100 0/32 Concentration of measure for Coulomb gases Idea of proof of Coulomb transport inequality Potential: if U µ (x) := g µ(x) then U µ (x) = c d µ Electric field: U µ (x). Carré du champ : U µ 2 Integration by parts + Schwarz s inequality in R d and L 2 c d f (x)(µ ν)(dx) = f (x) U µ ν (x)dx f (x) U µ ν (x) dx f Lip D + U µ ν (x) dx f Lip ( D + U µ ν (x) 2 dx) 1/2.

101 1/32 Concentration of measure for Coulomb gases Idea... Continued Again by integration by parts U µ ν (x) 2 dx = = c d = c d E(µ ν). U µ ν (x) U µ ν (x)dx U µ ν (x)(µ ν)(dx) Finally W 1 (µ, ν) 2 D + c d E(µ ν).

102 32/32 Concentration of measure for Coulomb gases Idea of proof of concentration dp N V,β (W 1(µ N, µ V ) r) = 1 ZV N e β 2 E (µ N ) dx.,β W 1 (µ N,µ V ) r Normalizing constant 1 Z N V,β exp { N 2 β 2 E V (µ V ) N ( )} β 2 E(µ V ) S(µ V ).

103 32/32 Concentration of measure for Coulomb gases Idea of proof of concentration dp N V,β (W 1(µ N, µ V ) r) = 1 ZV N e β 2 E (µ N ) dx.,β W 1 (µ N,µ V ) r Normalizing constant 1 Z N V,β exp { N 2 β 2 E V (µ V ) N ( )} β 2 E(µ V ) S(µ V ). Regularization: g superharmonic, µ (ε) N := µ N λ ε, E (µ N ) N 2 E V (µ (ε) N ) + NE(λ ε) + N N (V λ ε V )(x i ). i=1

104 Concentration of measure for Coulomb gases Idea of proof of concentration dp N V,β (W 1(µ N, µ V ) r) = 1 ZV N e β 2 E (µ N ) dx.,β W 1 (µ N,µ V ) r Normalizing constant 1 Z N V,β exp { N 2 β 2 E V (µ V ) N ( )} β 2 E(µ V ) S(µ V ). 32/32 Regularization: g superharmonic, µ (ε) N := µ N λ ε, E (µ N ) N 2 E V (µ (ε) N ) + NE(λ ε) + N N (V λ ε V )(x i ). Coulomb transport E V (µ (ε) N ) + E V (µ V ) 1 C W2 1 (µ(ε) N, µ V ). i=1

Concentration for Coulomb gases and Coulomb transport inequalities

Concentration for Coulomb gases and Coulomb transport inequalities Concentration for Coulomb gases and Coulomb transport inequalities Mylène Maïda U. Lille, Laboratoire Paul Painlevé Joint work with Djalil Chafaï and Adrien Hardy U. Paris-Dauphine and U. Lille ICERM,

More information

arxiv: v3 [math-ph] 29 Aug 2017

arxiv: v3 [math-ph] 29 Aug 2017 COCETRATIO FOR COULOMB GASES AD COULOMB TRASPORT IEQUALITIES DJALIL CHAFAÏ, ADRIE HARDY, AD MYLÈE MAÏDA arxiv:1610.00980v3 [math-ph] 29 Aug 2017 Abstract. We study the non-asymptotic behavior of Coulomb

More information

Convergence of spectral measures and eigenvalue rigidity

Convergence of spectral measures and eigenvalue rigidity Convergence of spectral measures and eigenvalue rigidity Elizabeth Meckes Case Western Reserve University ICERM, March 1, 2018 Macroscopic scale: the empirical spectral measure Macroscopic scale: the empirical

More information

Microscopic behavior for β-ensembles: an energy approach

Microscopic behavior for β-ensembles: an energy approach Microscopic behavior for β-ensembles: an energy approach Thomas Leblé (joint with/under the supervision of) Sylvia Serfaty Université Paris 6 BIRS workshop, 14 April 2016 Thomas Leblé (Université Paris

More information

DLR equations for the Sineβ process and applications

DLR equations for the Sineβ process and applications DLR equations for the Sineβ process and applications Thomas Leble (Courant Institute - NYU) Columbia University - 09/28/2018 Joint work with D. Dereudre, A. Hardy, M. Maı da (Universite Lille 1) Log-gases

More information

Concentration inequalities: basics and some new challenges

Concentration inequalities: basics and some new challenges Concentration inequalities: basics and some new challenges M. Ledoux University of Toulouse, France & Institut Universitaire de France Measure concentration geometric functional analysis, probability theory,

More information

Eigenvalue variance bounds for Wigner and covariance random matrices

Eigenvalue variance bounds for Wigner and covariance random matrices Eigenvalue variance bounds for Wigner and covariance random matrices S. Dallaporta University of Toulouse, France Abstract. This work is concerned with finite range bounds on the variance of individual

More information

Concentration Inequalities for Random Matrices

Concentration Inequalities for Random Matrices Concentration Inequalities for Random Matrices M. Ledoux Institut de Mathématiques de Toulouse, France exponential tail inequalities classical theme in probability and statistics quantify the asymptotic

More information

Ferromagnets and the classical Heisenberg model. Kay Kirkpatrick, UIUC

Ferromagnets and the classical Heisenberg model. Kay Kirkpatrick, UIUC Ferromagnets and the classical Heisenberg model Kay Kirkpatrick, UIUC Ferromagnets and the classical Heisenberg model: asymptotics for a mean-field phase transition Kay Kirkpatrick, Urbana-Champaign June

More information

Random regular digraphs: singularity and spectrum

Random regular digraphs: singularity and spectrum Random regular digraphs: singularity and spectrum Nick Cook, UCLA Probability Seminar, Stanford University November 2, 2015 Universality Circular law Singularity probability Talk outline 1 Universality

More information

Stein s method, logarithmic Sobolev and transport inequalities

Stein s method, logarithmic Sobolev and transport inequalities Stein s method, logarithmic Sobolev and transport inequalities M. Ledoux University of Toulouse, France and Institut Universitaire de France Stein s method, logarithmic Sobolev and transport inequalities

More information

Refined Bounds on the Empirical Distribution of Good Channel Codes via Concentration Inequalities

Refined Bounds on the Empirical Distribution of Good Channel Codes via Concentration Inequalities Refined Bounds on the Empirical Distribution of Good Channel Codes via Concentration Inequalities Maxim Raginsky and Igal Sason ISIT 2013, Istanbul, Turkey Capacity-Achieving Channel Codes The set-up DMC

More information

Fluctuations of random tilings and discrete Beta-ensembles

Fluctuations of random tilings and discrete Beta-ensembles Fluctuations of random tilings and discrete Beta-ensembles Alice Guionnet CRS (E S Lyon) Workshop in geometric functional analysis, MSRI, nov. 13 2017 Joint work with A. Borodin, G. Borot, V. Gorin, J.Huang

More information

Ferromagnets and superconductors. Kay Kirkpatrick, UIUC

Ferromagnets and superconductors. Kay Kirkpatrick, UIUC Ferromagnets and superconductors Kay Kirkpatrick, UIUC Ferromagnet and superconductor models: Phase transitions and asymptotics Kay Kirkpatrick, Urbana-Champaign October 2012 Ferromagnet and superconductor

More information

Central Limit Theorems for linear statistics for Biorthogonal Ensembles

Central Limit Theorems for linear statistics for Biorthogonal Ensembles Central Limit Theorems for linear statistics for Biorthogonal Ensembles Maurice Duits, Stockholm University Based on joint work with Jonathan Breuer (HUJI) Princeton, April 2, 2014 M. Duits (SU) CLT s

More information

Free Talagrand Inequality, a Simple Proof. Ionel Popescu. Northwestern University & IMAR

Free Talagrand Inequality, a Simple Proof. Ionel Popescu. Northwestern University & IMAR Free Talagrand Inequality, a Simple Proof Ionel Popescu Northwestern University & IMAR A Joke If F : [0, 1] Ris a smooth convex function such that F(0)=F (0)=0, then F(t) 0 for any t [0, 1]. Proof. F is

More information

Rigidity of the 3D hierarchical Coulomb gas. Sourav Chatterjee

Rigidity of the 3D hierarchical Coulomb gas. Sourav Chatterjee Rigidity of point processes Let P be a Poisson point process of intensity n in R d, and let A R d be a set of nonzero volume. Let N(A) := A P. Then E(N(A)) = Var(N(A)) = vol(a)n. Thus, N(A) has fluctuations

More information

Exponential tail inequalities for eigenvalues of random matrices

Exponential tail inequalities for eigenvalues of random matrices Exponential tail inequalities for eigenvalues of random matrices M. Ledoux Institut de Mathématiques de Toulouse, France exponential tail inequalities classical theme in probability and statistics quantify

More information

Free Probability and Random Matrices: from isomorphisms to universality

Free Probability and Random Matrices: from isomorphisms to universality Free Probability and Random Matrices: from isomorphisms to universality Alice Guionnet MIT TexAMP, November 21, 2014 Joint works with F. Bekerman, Y. Dabrowski, A.Figalli, E. Maurel-Segala, J. Novak, D.

More information

Upper Bound for Intermediate Singular Values of Random Sub-Gaussian Matrices 1

Upper Bound for Intermediate Singular Values of Random Sub-Gaussian Matrices 1 Upper Bound for Intermediate Singular Values of Random Sub-Gaussian Matrices 1 Feng Wei 2 University of Michigan July 29, 2016 1 This presentation is based a project under the supervision of M. Rudelson.

More information

RENORMALIZATION OF DYSON S VECTOR-VALUED HIERARCHICAL MODEL AT LOW TEMPERATURES

RENORMALIZATION OF DYSON S VECTOR-VALUED HIERARCHICAL MODEL AT LOW TEMPERATURES RENORMALIZATION OF DYSON S VECTOR-VALUED HIERARCHICAL MODEL AT LOW TEMPERATURES P. M. Bleher (1) and P. Major (2) (1) Keldysh Institute of Applied Mathematics of the Soviet Academy of Sciences Moscow (2)

More information

Concentration inequalities and the entropy method

Concentration inequalities and the entropy method Concentration inequalities and the entropy method Gábor Lugosi ICREA and Pompeu Fabra University Barcelona what is concentration? We are interested in bounding random fluctuations of functions of many

More information

Large sample covariance matrices and the T 2 statistic

Large sample covariance matrices and the T 2 statistic Large sample covariance matrices and the T 2 statistic EURANDOM, the Netherlands Joint work with W. Zhou Outline 1 2 Basic setting Let {X ij }, i, j =, be i.i.d. r.v. Write n s j = (X 1j,, X pj ) T and

More information

NEW FUNCTIONAL INEQUALITIES

NEW FUNCTIONAL INEQUALITIES 1 / 29 NEW FUNCTIONAL INEQUALITIES VIA STEIN S METHOD Giovanni Peccati (Luxembourg University) IMA, Minneapolis: April 28, 2015 2 / 29 INTRODUCTION Based on two joint works: (1) Nourdin, Peccati and Swan

More information

Semicircle law on short scales and delocalization for Wigner random matrices

Semicircle law on short scales and delocalization for Wigner random matrices Semicircle law on short scales and delocalization for Wigner random matrices László Erdős University of Munich Weizmann Institute, December 2007 Joint work with H.T. Yau (Harvard), B. Schlein (Munich)

More information

Concentration, self-bounding functions

Concentration, self-bounding functions Concentration, self-bounding functions S. Boucheron 1 and G. Lugosi 2 and P. Massart 3 1 Laboratoire de Probabilités et Modèles Aléatoires Université Paris-Diderot 2 Economics University Pompeu Fabra 3

More information

Local semicircle law, Wegner estimate and level repulsion for Wigner random matrices

Local semicircle law, Wegner estimate and level repulsion for Wigner random matrices Local semicircle law, Wegner estimate and level repulsion for Wigner random matrices László Erdős University of Munich Oberwolfach, 2008 Dec Joint work with H.T. Yau (Harvard), B. Schlein (Cambrigde) Goal:

More information

COMPLEX HERMITE POLYNOMIALS: FROM THE SEMI-CIRCULAR LAW TO THE CIRCULAR LAW

COMPLEX HERMITE POLYNOMIALS: FROM THE SEMI-CIRCULAR LAW TO THE CIRCULAR LAW Serials Publications www.serialspublications.com OMPLEX HERMITE POLYOMIALS: FROM THE SEMI-IRULAR LAW TO THE IRULAR LAW MIHEL LEDOUX Abstract. We study asymptotics of orthogonal polynomial measures of the

More information

Asymptotic stability of an evolutionary nonlinear Boltzmann-type equation

Asymptotic stability of an evolutionary nonlinear Boltzmann-type equation Acta Polytechnica Hungarica Vol. 14, No. 5, 217 Asymptotic stability of an evolutionary nonlinear Boltzmann-type equation Roksana Brodnicka, Henryk Gacki Institute of Mathematics, University of Silesia

More information

Local microscopic behavior for 2D Coulomb gases

Local microscopic behavior for 2D Coulomb gases Local microscopic behavior for 2D Coulomb gases Thomas Leblé October 0, 206 Abstract The study of two-dimensional Coulomb gases lies at the interface of statistical physics and non-hermitian random matrix

More information

Markov operators, classical orthogonal polynomial ensembles, and random matrices

Markov operators, classical orthogonal polynomial ensembles, and random matrices Markov operators, classical orthogonal polynomial ensembles, and random matrices M. Ledoux, Institut de Mathématiques de Toulouse, France 5ecm Amsterdam, July 2008 recent study of random matrix and random

More information

Free Entropy for Free Gibbs Laws Given by Convex Potentials

Free Entropy for Free Gibbs Laws Given by Convex Potentials Free Entropy for Free Gibbs Laws Given by Convex Potentials David A. Jekel University of California, Los Angeles Young Mathematicians in C -algebras, August 2018 David A. Jekel (UCLA) Free Entropy YMC

More information

SEPARABILITY AND COMPLETENESS FOR THE WASSERSTEIN DISTANCE

SEPARABILITY AND COMPLETENESS FOR THE WASSERSTEIN DISTANCE SEPARABILITY AND COMPLETENESS FOR THE WASSERSTEIN DISTANCE FRANÇOIS BOLLEY Abstract. In this note we prove in an elementary way that the Wasserstein distances, which play a basic role in optimal transportation

More information

Free energy estimates for the two-dimensional Keller-Segel model

Free energy estimates for the two-dimensional Keller-Segel model Free energy estimates for the two-dimensional Keller-Segel model dolbeaul@ceremade.dauphine.fr CEREMADE CNRS & Université Paris-Dauphine in collaboration with A. Blanchet (CERMICS, ENPC & Ceremade) & B.

More information

Stein s Method: Distributional Approximation and Concentration of Measure

Stein s Method: Distributional Approximation and Concentration of Measure Stein s Method: Distributional Approximation and Concentration of Measure Larry Goldstein University of Southern California 36 th Midwest Probability Colloquium, 2014 Concentration of Measure Distributional

More information

Regularity of local minimizers of the interaction energy via obstacle problems

Regularity of local minimizers of the interaction energy via obstacle problems Regularity of local minimizers of the interaction energy via obstacle problems J. A. Carrillo, M. G. Delgadino, A. Mellet September 22, 2014 Abstract The repulsion strength at the origin for repulsive/attractive

More information

Large deviations for random projections of l p balls

Large deviations for random projections of l p balls 1/32 Large deviations for random projections of l p balls Nina Gantert CRM, september 5, 2016 Goal: Understanding random projections of high-dimensional convex sets. 2/32 2/32 Outline Goal: Understanding

More information

Stability results for Logarithmic Sobolev inequality

Stability results for Logarithmic Sobolev inequality Stability results for Logarithmic Sobolev inequality Daesung Kim (joint work with Emanuel Indrei) Department of Mathematics Purdue University September 20, 2017 Daesung Kim (Purdue) Stability for LSI Probability

More information

Invariant measures for iterated function systems

Invariant measures for iterated function systems ANNALES POLONICI MATHEMATICI LXXV.1(2000) Invariant measures for iterated function systems by Tomasz Szarek (Katowice and Rzeszów) Abstract. A new criterion for the existence of an invariant distribution

More information

Concentration inequalities and tail bounds

Concentration inequalities and tail bounds Concentration inequalities and tail bounds John Duchi Outline I Basics and motivation 1 Law of large numbers 2 Markov inequality 3 Cherno bounds II Sub-Gaussian random variables 1 Definitions 2 Examples

More information

A Note On Large Deviation Theory and Beyond

A Note On Large Deviation Theory and Beyond A Note On Large Deviation Theory and Beyond Jin Feng In this set of notes, we will develop and explain a whole mathematical theory which can be highly summarized through one simple observation ( ) lim

More information

Dynamics of a planar Coulomb gas

Dynamics of a planar Coulomb gas Dynamics of a planar Coulomb gas François Bolley, Djalil Chafaï, Joaquin Fontbona To cite this version: François Bolley, Djalil Chafaï, Joaquin Fontbona. Dynamics of a planar Coulomb gas. Minor revision

More information

Fluctuations of random tilings and discrete Beta-ensembles

Fluctuations of random tilings and discrete Beta-ensembles Fluctuations of random tilings and discrete Beta-ensembles Alice Guionnet CRS (E S Lyon) Advances in Mathematics and Theoretical Physics, Roma, 2017 Joint work with A. Borodin, G. Borot, V. Gorin, J.Huang

More information

A Note on the Central Limit Theorem for the Eigenvalue Counting Function of Wigner and Covariance Matrices

A Note on the Central Limit Theorem for the Eigenvalue Counting Function of Wigner and Covariance Matrices A Note on the Central Limit Theorem for the Eigenvalue Counting Function of Wigner and Covariance Matrices S. Dallaporta University of Toulouse, France Abstract. This note presents some central limit theorems

More information

PCA with random noise. Van Ha Vu. Department of Mathematics Yale University

PCA with random noise. Van Ha Vu. Department of Mathematics Yale University PCA with random noise Van Ha Vu Department of Mathematics Yale University An important problem that appears in various areas of applied mathematics (in particular statistics, computer science and numerical

More information

arxiv: v4 [math-ph] 28 Feb 2018

arxiv: v4 [math-ph] 28 Feb 2018 FLUCTUATIOS OF TWO DIMESIOAL COULOMB GASES THOMAS LEBLÉ AD SYLVIA SERFATY arxiv:1609.08088v4 [math-ph] 28 Feb 2018 Abstract. We prove a Central Limit Theorem for the linear statistics of two-dimensional

More information

Free probabilities and the large N limit, IV. Loop equations and all-order asymptotic expansions... Gaëtan Borot

Free probabilities and the large N limit, IV. Loop equations and all-order asymptotic expansions... Gaëtan Borot Free probabilities and the large N limit, IV March 27th 2014 Loop equations and all-order asymptotic expansions... Gaëtan Borot MPIM Bonn & MIT based on joint works with Alice Guionnet, MIT Karol Kozlowski,

More information

2D Electrostatics and the Density of Quantum Fluids

2D Electrostatics and the Density of Quantum Fluids 2D Electrostatics and the Density of Quantum Fluids Jakob Yngvason, University of Vienna with Elliott H. Lieb, Princeton University and Nicolas Rougerie, University of Grenoble Yerevan, September 5, 2016

More information

Logarithmic, Coulomb and Riesz energy of point processes

Logarithmic, Coulomb and Riesz energy of point processes Logarithmic, Coulomb and Riesz energy of point processes Thomas Leblé September 7, 25 Abstract We define a notion of logarithmic, Coulomb and Riesz interactions in any dimension for random systems of infinite

More information

Random Matrix: From Wigner to Quantum Chaos

Random Matrix: From Wigner to Quantum Chaos Random Matrix: From Wigner to Quantum Chaos Horng-Tzer Yau Harvard University Joint work with P. Bourgade, L. Erdős, B. Schlein and J. Yin 1 Perhaps I am now too courageous when I try to guess the distribution

More information

Invertibility of random matrices

Invertibility of random matrices University of Michigan February 2011, Princeton University Origins of Random Matrix Theory Statistics (Wishart matrices) PCA of a multivariate Gaussian distribution. [Gaël Varoquaux s blog gael-varoquaux.info]

More information

Lecture I: Asymptotics for large GUE random matrices

Lecture I: Asymptotics for large GUE random matrices Lecture I: Asymptotics for large GUE random matrices Steen Thorbjørnsen, University of Aarhus andom Matrices Definition. Let (Ω, F, P) be a probability space and let n be a positive integer. Then a random

More information

A note on the convex infimum convolution inequality

A note on the convex infimum convolution inequality A note on the convex infimum convolution inequality Naomi Feldheim, Arnaud Marsiglietti, Piotr Nayar, Jing Wang Abstract We characterize the symmetric measures which satisfy the one dimensional convex

More information

Uniform concentration inequalities, martingales, Rademacher complexity and symmetrization

Uniform concentration inequalities, martingales, Rademacher complexity and symmetrization Uniform concentration inequalities, martingales, Rademacher complexity and symmetrization John Duchi Outline I Motivation 1 Uniform laws of large numbers 2 Loss minimization and data dependence II Uniform

More information

Approximations of displacement interpolations by entropic interpolations

Approximations of displacement interpolations by entropic interpolations Approximations of displacement interpolations by entropic interpolations Christian Léonard Université Paris Ouest Mokaplan 10 décembre 2015 Interpolations in P(X ) X : Riemannian manifold (state space)

More information

1 Intro to RMT (Gene)

1 Intro to RMT (Gene) M705 Spring 2013 Summary for Week 2 1 Intro to RMT (Gene) (Also see the Anderson - Guionnet - Zeitouni book, pp.6-11(?) ) We start with two independent families of R.V.s, {Z i,j } 1 i

More information

Universality for random matrices and log-gases

Universality for random matrices and log-gases Universality for random matrices and log-gases László Erdős IST, Austria Ludwig-Maximilians-Universität, Munich, Germany Encounters Between Discrete and Continuous Mathematics Eötvös Loránd University,

More information

Heat kernels of some Schrödinger operators

Heat kernels of some Schrödinger operators Heat kernels of some Schrödinger operators Alexander Grigor yan Tsinghua University 28 September 2016 Consider an elliptic Schrödinger operator H = Δ + Φ, where Δ = n 2 i=1 is the Laplace operator in R

More information

Displacement convexity of the relative entropy in the discrete h

Displacement convexity of the relative entropy in the discrete h Displacement convexity of the relative entropy in the discrete hypercube LAMA Université Paris Est Marne-la-Vallée Phenomena in high dimensions in geometric analysis, random matrices, and computational

More information

Logarithmic Sobolev Inequalities

Logarithmic Sobolev Inequalities Logarithmic Sobolev Inequalities M. Ledoux Institut de Mathématiques de Toulouse, France logarithmic Sobolev inequalities what they are, some history analytic, geometric, optimal transportation proofs

More information

Random Matrices: Beyond Wigner and Marchenko-Pastur Laws

Random Matrices: Beyond Wigner and Marchenko-Pastur Laws Random Matrices: Beyond Wigner and Marchenko-Pastur Laws Nathan Noiry Modal X, Université Paris Nanterre May 3, 2018 Wigner s Matrices Wishart s Matrices Generalizations Wigner s Matrices ij, (n, i, j

More information

The Dynamics of Learning: A Random Matrix Approach

The Dynamics of Learning: A Random Matrix Approach The Dynamics of Learning: A Random Matrix Approach ICML 2018, Stockholm, Sweden Zhenyu Liao, Romain Couillet L2S, CentraleSupélec, Université Paris-Saclay, France GSTATS IDEX DataScience Chair, GIPSA-lab,

More information

Asymptotic results for empirical measures of weighted sums of independent random variables

Asymptotic results for empirical measures of weighted sums of independent random variables Asymptotic results for empirical measures of weighted sums of independent random variables B. Bercu and W. Bryc University Bordeaux 1, France Workshop on Limit Theorems, University Paris 1 Paris, January

More information

Stability of boundary measures

Stability of boundary measures Stability of boundary measures F. Chazal D. Cohen-Steiner Q. Mérigot INRIA Saclay - Ile de France LIX, January 2008 Point cloud geometry Given a set of points sampled near an unknown shape, can we infer

More information

Duality of multiparameter Hardy spaces H p on spaces of homogeneous type

Duality of multiparameter Hardy spaces H p on spaces of homogeneous type Duality of multiparameter Hardy spaces H p on spaces of homogeneous type Yongsheng Han, Ji Li, and Guozhen Lu Department of Mathematics Vanderbilt University Nashville, TN Internet Analysis Seminar 2012

More information

Asymptotics for posterior hazards

Asymptotics for posterior hazards Asymptotics for posterior hazards Pierpaolo De Blasi University of Turin 10th August 2007, BNR Workshop, Isaac Newton Intitute, Cambridge, UK Joint work with Giovanni Peccati (Université Paris VI) and

More information

Higher Dimensional Coulomb Gases and Renormalized Energy Functionals

Higher Dimensional Coulomb Gases and Renormalized Energy Functionals Higher Dimensional Coulomb Gases and Renormalized Energy Functionals N. Rougerie and S. Serfaty July 9, 2013 Abstract We consider a classical system of n charged particles in an external confining potential,

More information

Class 2 & 3 Overfitting & Regularization

Class 2 & 3 Overfitting & Regularization Class 2 & 3 Overfitting & Regularization Carlo Ciliberto Department of Computer Science, UCL October 18, 2017 Last Class The goal of Statistical Learning Theory is to find a good estimator f n : X Y, approximating

More information

Practical conditions on Markov chains for weak convergence of tail empirical processes

Practical conditions on Markov chains for weak convergence of tail empirical processes Practical conditions on Markov chains for weak convergence of tail empirical processes Olivier Wintenberger University of Copenhagen and Paris VI Joint work with Rafa l Kulik and Philippe Soulier Toronto,

More information

Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio ( )

Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio ( ) Mathematical Methods for Neurosciences. ENS - Master MVA Paris 6 - Master Maths-Bio (2014-2015) Etienne Tanré - Olivier Faugeras INRIA - Team Tosca October 22nd, 2014 E. Tanré (INRIA - Team Tosca) Mathematical

More information

Contraction Principles some applications

Contraction Principles some applications Contraction Principles some applications Fabrice Gamboa (Institut de Mathématiques de Toulouse) 7th of June 2017 Christian and Patrick 59th Birthday Overview Appetizer : Christian and Patrick secret lives

More information

Probabilistic Methods in Asymptotic Geometric Analysis.

Probabilistic Methods in Asymptotic Geometric Analysis. Probabilistic Methods in Asymptotic Geometric Analysis. C. Hugo Jiménez PUC-RIO, Brazil September 21st, 2016. Colmea. RJ 1 Origin 2 Normed Spaces 3 Distribution of volume of high dimensional convex bodies

More information

STAT 200C: High-dimensional Statistics

STAT 200C: High-dimensional Statistics STAT 200C: High-dimensional Statistics Arash A. Amini May 30, 2018 1 / 59 Classical case: n d. Asymptotic assumption: d is fixed and n. Basic tools: LLN and CLT. High-dimensional setting: n d, e.g. n/d

More information

Joint distribution optimal transportation for domain adaptation

Joint distribution optimal transportation for domain adaptation Joint distribution optimal transportation for domain adaptation Changhuang Wan Mechanical and Aerospace Engineering Department The Ohio State University March 8 th, 2018 Joint distribution optimal transportation

More information

LARGE DEVIATION PRINCIPLES FOR HYPERSINGULAR RIESZ GASES

LARGE DEVIATION PRINCIPLES FOR HYPERSINGULAR RIESZ GASES LARGE DEVIATION PRINCIPLES FOR HYPERSINGULAR RIESZ GASES DOUGLAS P. HARDIN, THOMAS LEBLÉ, EDWARD B. SAFF, SYLVIA SERFATY Abstract. We study N-particle systems in R d whose interactions are governed by

More information

Nishant Gurnani. GAN Reading Group. April 14th, / 107

Nishant Gurnani. GAN Reading Group. April 14th, / 107 Nishant Gurnani GAN Reading Group April 14th, 2017 1 / 107 Why are these Papers Important? 2 / 107 Why are these Papers Important? Recently a large number of GAN frameworks have been proposed - BGAN, LSGAN,

More information

Assessing the dependence of high-dimensional time series via sample autocovariances and correlations

Assessing the dependence of high-dimensional time series via sample autocovariances and correlations Assessing the dependence of high-dimensional time series via sample autocovariances and correlations Johannes Heiny University of Aarhus Joint work with Thomas Mikosch (Copenhagen), Richard Davis (Columbia),

More information

Inhomogeneous circular laws for random matrices with non-identically distributed entries

Inhomogeneous circular laws for random matrices with non-identically distributed entries Inhomogeneous circular laws for random matrices with non-identically distributed entries Nick Cook with Walid Hachem (Telecom ParisTech), Jamal Najim (Paris-Est) and David Renfrew (SUNY Binghamton/IST

More information

The circular law. Lewis Memorial Lecture / DIMACS minicourse March 19, Terence Tao (UCLA)

The circular law. Lewis Memorial Lecture / DIMACS minicourse March 19, Terence Tao (UCLA) The circular law Lewis Memorial Lecture / DIMACS minicourse March 19, 2008 Terence Tao (UCLA) 1 Eigenvalue distributions Let M = (a ij ) 1 i n;1 j n be a square matrix. Then one has n (generalised) eigenvalues

More information

Nonparametric regression with martingale increment errors

Nonparametric regression with martingale increment errors S. Gaïffas (LSTA - Paris 6) joint work with S. Delattre (LPMA - Paris 7) work in progress Motivations Some facts: Theoretical study of statistical algorithms requires stationary and ergodicity. Concentration

More information

Integral Representation Formula, Boundary Integral Operators and Calderón projection

Integral Representation Formula, Boundary Integral Operators and Calderón projection Integral Representation Formula, Boundary Integral Operators and Calderón projection Seminar BEM on Wave Scattering Franziska Weber ETH Zürich October 22, 2010 Outline Integral Representation Formula Newton

More information

Heat Flows, Geometric and Functional Inequalities

Heat Flows, Geometric and Functional Inequalities Heat Flows, Geometric and Functional Inequalities M. Ledoux Institut de Mathématiques de Toulouse, France heat flow and semigroup interpolations Duhamel formula (19th century) pde, probability, dynamics

More information

The Matrix Dyson Equation in random matrix theory

The Matrix Dyson Equation in random matrix theory The Matrix Dyson Equation in random matrix theory László Erdős IST, Austria Mathematical Physics seminar University of Bristol, Feb 3, 207 Joint work with O. Ajanki, T. Krüger Partially supported by ERC

More information

Convergence rates in weighted L 1 spaces of kernel density estimators for linear processes

Convergence rates in weighted L 1 spaces of kernel density estimators for linear processes Alea 4, 117 129 (2008) Convergence rates in weighted L 1 spaces of kernel density estimators for linear processes Anton Schick and Wolfgang Wefelmeyer Anton Schick, Department of Mathematical Sciences,

More information

Contents 1. Introduction 1 2. Main results 3 3. Proof of the main inequalities 7 4. Application to random dynamical systems 11 References 16

Contents 1. Introduction 1 2. Main results 3 3. Proof of the main inequalities 7 4. Application to random dynamical systems 11 References 16 WEIGHTED CSISZÁR-KULLBACK-PINSKER INEQUALITIES AND APPLICATIONS TO TRANSPORTATION INEQUALITIES FRANÇOIS BOLLEY AND CÉDRIC VILLANI Abstract. We strengthen the usual Csiszár-Kullback-Pinsker inequality by

More information

Optimal transportation and optimal control in a finite horizon framework

Optimal transportation and optimal control in a finite horizon framework Optimal transportation and optimal control in a finite horizon framework Guillaume Carlier and Aimé Lachapelle Université Paris-Dauphine, CEREMADE July 2008 1 MOTIVATIONS - A commitment problem (1) Micro

More information

On a Class of Multidimensional Optimal Transportation Problems

On a Class of Multidimensional Optimal Transportation Problems Journal of Convex Analysis Volume 10 (2003), No. 2, 517 529 On a Class of Multidimensional Optimal Transportation Problems G. Carlier Université Bordeaux 1, MAB, UMR CNRS 5466, France and Université Bordeaux

More information

Asymptotic results for empirical measures of weighted sums of independent random variables

Asymptotic results for empirical measures of weighted sums of independent random variables Asymptotic results for empirical measures of weighted sums of independent random variables B. Bercu and W. Bryc University Bordeaux 1, France Seminario di Probabilità e Statistica Matematica Sapienza Università

More information

(somewhat) expanded version of the note in C. R. Acad. Sci. Paris 340, (2005). A (ONE-DIMENSIONAL) FREE BRUNN-MINKOWSKI INEQUALITY

(somewhat) expanded version of the note in C. R. Acad. Sci. Paris 340, (2005). A (ONE-DIMENSIONAL) FREE BRUNN-MINKOWSKI INEQUALITY (somewhat expanded version of the note in C. R. Acad. Sci. Paris 340, 30 304 (2005. A (ONE-DIMENSIONAL FREE BRUNN-MINKOWSKI INEQUALITY M. Ledoux University of Toulouse, France Abstract. We present a one-dimensional

More information

The norm of polynomials in large random matrices

The norm of polynomials in large random matrices The norm of polynomials in large random matrices Camille Mâle, École Normale Supérieure de Lyon, Ph.D. Student under the direction of Alice Guionnet. with a significant contribution by Dimitri Shlyakhtenko.

More information

A Lévy-Fokker-Planck equation: entropies and convergence to equilibrium

A Lévy-Fokker-Planck equation: entropies and convergence to equilibrium 1/ 22 A Lévy-Fokker-Planck equation: entropies and convergence to equilibrium I. Gentil CEREMADE, Université Paris-Dauphine International Conference on stochastic Analysis and Applications Hammamet, Tunisia,

More information

Finite-dimensional spaces. C n is the space of n-tuples x = (x 1,..., x n ) of complex numbers. It is a Hilbert space with the inner product

Finite-dimensional spaces. C n is the space of n-tuples x = (x 1,..., x n ) of complex numbers. It is a Hilbert space with the inner product Chapter 4 Hilbert Spaces 4.1 Inner Product Spaces Inner Product Space. A complex vector space E is called an inner product space (or a pre-hilbert space, or a unitary space) if there is a mapping (, )

More information

Incompressibility Estimates in the Laughlin Phase

Incompressibility Estimates in the Laughlin Phase Incompressibility Estimates in the Laughlin Phase Jakob Yngvason, University of Vienna with Nicolas Rougerie, University of Grenoble București, July 2, 2014 Jakob Yngvason (Uni Vienna) Incompressibility

More information

Lecture No 1 Introduction to Diffusion equations The heat equat

Lecture No 1 Introduction to Diffusion equations The heat equat Lecture No 1 Introduction to Diffusion equations The heat equation Columbia University IAS summer program June, 2009 Outline of the lectures We will discuss some basic models of diffusion equations and

More information

The Moment Method; Convex Duality; and Large/Medium/Small Deviations

The Moment Method; Convex Duality; and Large/Medium/Small Deviations Stat 928: Statistical Learning Theory Lecture: 5 The Moment Method; Convex Duality; and Large/Medium/Small Deviations Instructor: Sham Kakade The Exponential Inequality and Convex Duality The exponential

More information

Weak quenched limiting distributions of a one-dimensional random walk in a random environment

Weak quenched limiting distributions of a one-dimensional random walk in a random environment Weak quenched limiting distributions of a one-dimensional random walk in a random environment Jonathon Peterson Cornell University Department of Mathematics Joint work with Gennady Samorodnitsky September

More information

Econ Lecture 3. Outline. 1. Metric Spaces and Normed Spaces 2. Convergence of Sequences in Metric Spaces 3. Sequences in R and R n

Econ Lecture 3. Outline. 1. Metric Spaces and Normed Spaces 2. Convergence of Sequences in Metric Spaces 3. Sequences in R and R n Econ 204 2011 Lecture 3 Outline 1. Metric Spaces and Normed Spaces 2. Convergence of Sequences in Metric Spaces 3. Sequences in R and R n 1 Metric Spaces and Metrics Generalize distance and length notions

More information

Distance-Divergence Inequalities

Distance-Divergence Inequalities Distance-Divergence Inequalities Katalin Marton Alfréd Rényi Institute of Mathematics of the Hungarian Academy of Sciences Motivation To find a simple proof of the Blowing-up Lemma, proved by Ahlswede,

More information

Graduate Econometrics I: Asymptotic Theory

Graduate Econometrics I: Asymptotic Theory Graduate Econometrics I: Asymptotic Theory Yves Dominicy Université libre de Bruxelles Solvay Brussels School of Economics and Management ECARES Yves Dominicy Graduate Econometrics I: Asymptotic Theory

More information

Wigner s semicircle law

Wigner s semicircle law CHAPTER 2 Wigner s semicircle law 1. Wigner matrices Definition 12. A Wigner matrix is a random matrix X =(X i, j ) i, j n where (1) X i, j, i < j are i.i.d (real or complex valued). (2) X i,i, i n are

More information