Searcharxiv⌕ Search

arXiv subjects

Assaf Naor

Publications and source records attributed to Assaf Naor.

At least 37 records · Page 2Linked to original sources

Concentration of Markov chains with bounded moments

Let $\{W_t\}_{t=1}^{\infty}$ be a finite state stationary Markov chain, and suppose that $f$ is a real-valued function on the state space. If $f$ is bounded, then Gillman's expander Chernoff bound (1993) provides concentration estimates for the random variable $f(W_1)+\cdots+f(W_n)$ that depend on the spectral gap of the Markov chain and the assumed bound on $f$. Here we obtain analogous inequalities assuming only that the $q$'th moment of $f$ is bounded for some $q \geq 2$. Our proof relies on reasoning that differs substantially from the proofs of Gillman's theorem that are available in the literature, and it generalizes to yield dimension-independent bounds for mappings $f$ that take values in an $L_p(μ)$ for some $p\ge 2$, thus answering (even in the Hilbertian special case $p=2$) a question of Kargin (2007).

math.PR↗

Nonpositive curvature is not coarsely universal

We prove that not every metric space embeds coarsely into an Alexandrov space of nonpositive curvature. This answers a question of Gromov (1993) and is in contrast to the fact that any metric space embeds coarsely into an Alexandrov space of nonnegative curvature, as shown by Andoni, Naor and Neiman (2015). We establish this statement by proving that a metric space which is $q$-barycentric for some $q\in [1,\infty)$ has metric cotype $q$ with sharp scaling parameter. Our proof utilizes nonlinear (metric space-valued) martingale inequalities and yields sharp bounds even for some classical Banach spaces. This allows us to evaluate the bi-Lipschitz distortion of the $\ell_\infty$ grid $[m]_\infty^n=(\{1,\ldots,m\}^n,\|\cdot\|_\infty)$ into $\ell_q$ for all $q\in (2,\infty)$, from which we deduce the following discrete converse to the fact that $\ell_\infty^n$ embeds with distortion $O(1)$ into $\ell_q$ for $q=O(\log n)$. A rigidity theorem of Ribe (1976) implies that for every $n\in \mathbb{N}$ there exists $m\in \mathbb{N}$ such that if $[m]_\infty^n$ embeds into $\ell_q$ with distortion $O(1)$, then $q$ is necessarily at least a universal constant multiple of $\log n$. Ribe's theorem does not give an explicit upper bound on this $m$, but by the work of Bourgain (1987) it suffices to take $m=n$, and this was the previously best-known estimate for $m$. We show that the above discretization statement actually holds when $m$ is a universal constant.

math.MG↗

The Andoni--Krauthgamer--Razenshteyn characterization of sketchable norms fails for sketchable metrics

Andoni, Krauthgamer and Razenshteyn (AKR) proved (STOC 2015) that a finite-dimensional normed space $(X,\|\cdot\|_X)$ admits a $O(1)$ sketching algorithm (namely, with $O(1)$ sketch size and $O(1)$ approximation) if and only if for every $\varepsilon\in (0,1)$ there exist $α\geqslant 1$ and an embedding $f:X\to \ell_{1-\varepsilon}$ such that $\|x-y\|_X\leqslant \|f(x)-f(y)\|_{1-\varepsilon}\leqslant α\|x-y\|_X$ for all $x,y\in X$. The "if part" of this theorem follows from a sketching algorithm of Indyk (FOCS 2000). The contribution of AKR is therefore to demonstrate that the mere availability of a sketching algorithm implies the existence of the aforementioned geometric realization. Indyk's algorithm shows that the "if part" of the AKR characterization holds true for any metric space whatsoever, i.e., the existence of an embedding as above implies sketchability even when $X$ is not a normed space. Due to this, a natural question that AKR posed was whether the assumption that the underlying space is a normed space is needed for their characterization of sketchability. We resolve this question by proving that for arbitrarily large $n\in \mathbb{N}$ there is an $n$-point metric space $(M(n),d_{M(n)})$ which is $O(1)$-sketchable yet for every $\varepsilon\in (0,\frac12)$, if $α(n)\geqslant 1$ and $f_n:M(n)\to \ell_{1-\varepsilon}$ are such that $d_{M(n)}(x,y)\leqslant \|f_n(x)-f_n(y)\|_{1-\varepsilon}\leqslant α(n) d_{M(n)}(x,y)$ for all $x,y\in M(n)$, then necessarily $\lim_{n\to \infty} α(n)= \infty$.

cs.DS↗

Vertical perimeter versus horizontal perimeter

The discrete Heisenberg group $\mathbb{H}_{\mathbb{Z}}^{2k+1}$ is the group generated by $a_1,b_1,\ldots,a_k,b_k,c$, subject to the relations $[a_1,b_1]=\ldots=[a_k,b_k]=c$ and $[a_i,a_j]=[b_i,b_j]=[a_i,b_j]=[a_i,c]=[b_i,c]=1$ for every distinct $i,j\in \{1,\ldots,k\}$. Denote $S=\{a_1^{\pm 1},b_1^{\pm 1},\ldots,a_k^{\pm 1},b_k^{\pm 1}\}$. The horizontal boundary of $Ω\subset \mathbb{H}_{\mathbb{Z}}^{2k+1}$, denoted $\partial_{h}Ω$, is the set of all $(x,y)\in Ω\times (\mathbb{H}_{\mathbb{Z}}^{2k+1}\setminus Ω)$ such that $x^{-1}y\in S$. The horizontal perimeter of $Ω$ is $|\partial_{h}Ω|$. For $t\in \mathbb{N}$, define $\partial^t_{v} Ω$ to be the set of all $(x,y)\in Ω\times (\mathbb{H}_{\mathsf{Z}}^{2k+1}\setminus Ω)$ such that $x^{-1}y\in \{c^t,c^{-t}\}$. The vertical perimeter of $Ω$ is defined by $|\partial_{v}Ω|= \sqrt{\sum_{t=1}^\infty |\partial^t_{v}Ω|^2/t^2}$. It is shown here that if $k\ge 2$, then $|\partial_{v}Ω|\lesssim \frac{1}{k} |\partial_{h}Ω|$. The proof of this "vertical versus horizontal isoperimetric inequality" uses a new structural result that decomposes sets of finite perimeter in the Heisenberg group into pieces that admit an "intrinsic corona decomposition." This allows one to deduce an endpoint $W^{1,1}\to L_2(L_1)$ boundedness of a certain singular integral operator from a corresponding lower-dimensional $W^{1,2}\to L_2(L_2)$ boundedness. The above inequality has several applications, including that any embedding into $L_1$ of a ball of radius $n$ in the word metric on $\mathbb{H}_{\mathbb{Z}}^{5}$ incurs bi-Lipschitz distortion that is at least a constant multiple of $\sqrt{\log n}$. It follows that the integrality gap of the Goemans--Linial semidefinite program for the Sparsest Cut Problem on inputs of size $n$ is at least a constant multiple of $\sqrt{\log n}$.

math.MG↗

Impossibility of dimension reduction in the nuclear norm

Let $\mathsf{S}_1$ (the Schatten--von Neumann trace class) denote the Banach space of all compact linear operators $T:\ell_2\to \ell_2$ whose nuclear norm $\|T\|_{\mathsf{S}_1}=\sum_{j=1}^\inftyσ_j(T)$ is finite, where $\{σ_j(T)\}_{j=1}^\infty$ are the singular values of $T$. We prove that for arbitrarily large $n\in \mathbb{N}$ there exists a subset $\mathcal{C}\subseteq \mathsf{S}_1$ with $|\mathcal{C}|=n$ that cannot be embedded with bi-Lipschitz distortion $O(1)$ into any $n^{o(1)}$-dimensional linear subspace of $\mathsf{S}_1$. $\mathcal{C}$ is not even a $O(1)$-Lipschitz quotient of any subset of any $n^{o(1)}$-dimensional linear subspace of $\mathsf{S}_1$. Thus, $\mathsf{S}_1$ does not admit a dimension reduction result á la Johnson and Lindenstrauss (1984), which complements the work of Harrow, Montanaro and Short (2011) on the limitations of quantum dimension reduction under the assumption that the embedding into low dimensions is a quantum channel. Such a statement was previously known with $\mathsf{S}_1$ replaced by the Banach space $\ell_1$ of absolutely summable sequences via the work of Brinkman and Charikar (2003). In fact, the above set $\mathcal{C}$ can be taken to be the same set as the one that Brinkman and Charikar considered, viewed as a collection of diagonal matrices in $\mathsf{S}_1$. The challenge is to demonstrate that $\mathcal{C}$ cannot be faithfully realized in an arbitrary low-dimensional subspace of $\mathsf{S}_1$, while Brinkman and Charikar obtained such an assertion only for subspaces of $\mathsf{S}_1$ that consist of diagonal operators (i.e., subspaces of $\ell_1$). We establish this by proving that the Markov 2-convexity constant of any finite dimensional linear subspace $X$ of $\mathsf{S}_1$ is at most a universal constant multiple of $\sqrt{\log \mathrm{dim}(X)}$.

math.FA↗

A relation between finitary Lipschitz extension moduli

This short note contains an elementary observation in response to the recent posting arXiv:1707.06593v1, which studies the Lipschitz extension modulus to $n$ additional points. We bound this modulus in terms of the well-studied Lipschitz extension modulus from $n$ points to the entire ambient metric space, thus making it possible to quote the available literature to improve some of the bounds obtained in arXiv:1707.06593v1.

math.MG↗

Quantitative affine approximation for UMD targets

It is shown here that if $(Y,\|\cdot\|_Y)$ is a Banach space in which martingale differences are unconditional (a UMD Banach space) then there exists $c=c(Y)\in (0,\infty)$ with the following property. For every $n\in \mathbb{N}$ and $\varepsilon\in (0,1/2]$, if $(X,\|\cdot\|_X)$ is an $n$-dimensional normed space with unit ball $B_X$ and $f:B_X\to Y$ is a $1$-Lipschitz function then there exists an affine mapping $Λ:X\to Y$ and a sub-ball $B^*=y+ρB_X\subseteq B_X$ of radius $ρ\ge \exp(-(1/\varepsilon)^{cn})$ such that $\|f(x)-Λ(x)\|_Y\le \varepsilon ρ$ for all $x\in B^*$. This estimate on the macroscopic scale of affine approximability of vector-valued Lipschitz functions is an asymptotic improvement (as $n\to \infty$) over the best previously known bound even when $X$ is $\mathbb{R}^n$ equipped with the Euclidean norm and $Y$ is a Hilbert space.

math.FA↗

A spectral gap precludes low-dimensional embeddings

We prove that there is a universal constant $C>0$ with the following property. Suppose that $n\in \mathbb{N}$ and that $\mathsf{A}=(a_{ij})\in M_n(\mathbb{R})$ is a symmetric stochastic matrix. Denote the second-largest eigenvalue of $\mathsf{A}$ by $λ_2(\mathsf{A})$. Then for $\mathrm{\it any}$ finite-dimensional normed space $(X,\|\cdot\|)$ we have $$ \forall\, x_1,\ldots,x_n\in X,\qquad \mathrm{dim}(X)\ge \frac12 \exp\left(C\frac{1-λ_2(\mathsf{A})}{\sqrt{n}}\bigg(\frac{\sum_{i=1}^n\sum_{j=1}^n\|x_i-x_j\|^2}{\sum_{i=1}^n\sum_{j=1}^na_{ij}\|x_i-x_j\|^2}\bigg)^{\frac12}\right). $$ This implies that if an $n$-vertex $O(1)$-expander embeds with average distortion $D\ge 1$ into $X$, then necessarily $\mathrm{dim}(X)\gtrsim n^{c/D}$ for some universal constant $c>0$, thus improving over the previously best-known estimate $\mathrm{dim}(X)\gtrsim (\log n)^2/D^2$ of Linial, London and Rabinovich, strengthening a theorem of Matoušek, and answering a question of Andoni, Nikolov, Razenshteyn and Waingarten.

math.MG↗

Restricted invertibility revisited

Suppose that $m,n\in \mathbb{N}$ and that $A:\mathbb{R}^m\to \mathbb{R}^n$ is a linear operator. It is shown here that if $k,r\in \mathbb{N}$ satisfy $k<r\le \mathrm{\bf rank(A)}$ then there exists a subset $σ\subseteq \{1,\ldots,m\}$ with $|σ|=k$ such that the restriction of $A$ to $\mathbb{R}^σ\subseteq \mathbb{R}^m$ is invertible, and moreover the operator norm of the inverse $A^{-1}:A(\mathbb{R}^σ)\to \mathbb{R}^m$ is at most a constant multiple of the quantity $\sqrt{mr/((r-k)\sum_{i=r}^m \mathsf{s}_i(A)^2)}$, where $\mathsf{s}_1(A)\geqslant\ldots\geqslant \mathsf{s}_m(A)$ are the singular values of $A$. This improves over a series of works, starting from the seminal Bourgain--Tzafriri Restricted Invertibility Principle, through the works of Vershynin, Spielman--Srivastava and Marcus--Spielman--Srivastava. In particular, this directly implies an improved restricted invertibility principle in terms of Schatten--von Neumann norms.

math.FA↗

Heat flow and quantitative differentiation

For every Banach space $(Y,\|\cdot\|_Y)$ that admits an equivalent uniformly convex norm we prove that there exists $c=c(Y)\in (0,\infty)$ with the following property. Suppose that $n\in \mathbb{N}$ and that $X$ is an $n$-dimensional normed space with unit ball $B_X$. Then for every $1$-Lipschitz function $f:B_X\to Y$ and for every $\varepsilon\in (0,1/2]$ there exists a radius $r\ge\exp(-1/\varepsilon^{cn})$, a point $x\in B_X$ with $x+rB_X\subset B_X$, and an affine mapping $Λ:X\to Y$ such that $\|f(y)-Λ(y)\|_Y\le \varepsilon r$ for every $y\in x+rB_X$.

math.FA↗

Discrete Riesz transforms and sharp metric $X_p$ inequalities

$ \renewcommand{\subset}{\subseteq} \newcommand{\N}{\mathbb N} $For $p\in [2,\infty)$ the metric $X_p$ inequality with sharp scaling parameter is proven here to hold true in $L_p$. The geometric consequences of this result include the following sharp statements about embeddings of $L_q$ into $L_p$ when $2< q<p<\infty$: the maximal $θ\in (0,1]$ for which $L_q$ admits a bi-$θ$-Hölder embedding into $L_p$ equals $q/p$, and for $m,n\in \N$ the smallest possible bi-Lipschitz distortion of any embedding into $L_p$ of the grid $\{1,\ldots,m\}^n\subset \ell_q^n$ is bounded above and below by constant multiples (depending only on $p,q$) of the quantity $\min\{n^{(p-q)(q-2)/(q^2(p-2))}, m^{(q-2)/q}\}$.

math.MG↗

Metric ${X}_p$ inequalities

For every $p\in (0,\infty)$ we associate to every metric space $(X,d_X)$ a numerical invariant $\mathfrak{X}_p(X)\in [0,\infty]$ such that if $\mathfrak{X}_p(X)<\infty$ and a metric space $(Y,d_Y)$ admits a bi-Lipschitz embedding into $X$ then also $\mathfrak{X}_p(Y)<\infty$. We prove that if $p,q\in (2,\infty)$ satisfy $q<p$ then $\mathfrak{X}_p(L_p)<\infty$ yet $\mathfrak{X}_p(L_q)=\infty$. Thus our new bi-Lipschitz invariant certifies that $L_q$ does not admit a bi-Lipschitz embedding into $L_p$ when $2<q<p<\infty$. This completes the long-standing search for bi-Lipschitz invariants that serve as an obstruction to the embeddability of $L_p$ spaces into each other, the previously understood cases of which were metric notions of type and cotype, which however fail to certify the nonembeddability of $L_q$ into $L_p$ when $2<q<p<\infty$. Among the consequences of our results are new quantitative restrictions on the bi-Lipschitz embeddability into $L_p$ of snowflakes of $L_q$ and integer grids in $\ell_q^n$, for $2<q<p<\infty$. As a byproduct of our investigations, we also obtain results on the geometry of the Schatten $p$ trace class $S_p$ that are new even in the linear setting.

math.FA↗

Snowflake universality of Wasserstein spaces

For $p\in (1,\infty)$ let $\mathscr{P}_p(\mathbb{R}^3)$ denote the metric space of all $p$-integrable Borel probability measures on $\mathbb{R}^3$, equipped with the Wasserstein $p$ metric $\mathsf{W}_p$. We prove that for every $\varepsilon>0$, every $θ\in (0,1/p]$ and every finite metric space $(X,d_X)$, the metric space $(X,d_{X}^θ)$ embeds into $\mathscr{P}_p(\mathbb{R}^3)$ with distortion at most $1+\varepsilon$. We show that this is sharp when $p\in (1,2]$ in the sense that the exponent $1/p$ cannot be replaced by any larger number. In fact, for arbitrarily large $n\in \mathbb{N}$ there exists an $n$-point metric space $(X_n,d_n)$ such that for every $α\in (1/p,1]$ any embedding of the metric space $(X_n,d_n^α)$ into $\mathscr{P}_p(\mathbb{R}^3)$ incurs distortion that is at least a constant multiple of $(\log n)^{α-1/p}$. These statements establish that there exists an Alexandrov space of nonnegative curvature, namely $\mathscr{P}_{\! 2}(\mathbb{R}^3)$, with respect to which there does not exist a sequence of bounded degree expander graphs. It also follows that $\mathscr{P}_{\! 2}(\mathbb{R}^3)$ does not admit a uniform, coarse, or quasisymmetric embedding into any Banach space of nontrivial type. Links to several longstanding open questions in metric geometry are discussed, including the characterization of subsets of Alexandrov spaces, existence of expanders, the universality problem for $\mathscr{P}_{\! 2}(\mathbb{R}^k)$, and the metric cotype dichotomy problem.

math.MG↗

Uniform nonextendability from nets

It is shown that there exist Banach spaces $X,Y$, a $1$-net $\mathscr{N}$ of $X$ and a Lipschitz function $f:\mathscr{N}\to Y$ such that every $F:X\to Y$ that extends $f$ is not uniformly continuous.

math.MG↗

On Lipschitz extension from finite subsets

We prove that for every $n\in \mathbb{N}$ there exists a metric space $(X,d_X)$, an $n$-point subset $S\subseteq X$, a Banach space $(Z,\|\cdot\|_Z)$ and a $1$-Lipschitz function $f:S\to Z$ such that the Lipschitz constant of every function $F:X\to Z$ that extends $f$ is at least a constant multiple of $\sqrt{\log n}$. This improves a bound of Johnson and Lindenstrauss. We also obtain the following quantitative counterpart to a classical extension theorem of Minty. For every $α\in (1/2,1]$ and $n\in \mathbb{N}$ there exists a metric space $(X,d_X)$, an $n$-point subset $S\subseteq X$ and a function $f:S\to \ell_2$ that is $α$-Hölder with constant $1$, yet the $α$-Hölder constant of any $F:X\to \ell_2$ that extends $f$ satisfies $$ \|F\|_{\mathrm{Lip}(α)}\gtrsim (\log n)^{\frac{2α-1}{4α}}+\left(\frac{\log n}{\log\log n}\right)^{α^2-\frac12}. $$ We formulate a conjecture whose positive solution would strengthen Ball's nonlinear Maurey extension theorem, serving as a far-reaching nonlinear version of a theorem of König, Retherford and Tomczak-Jaegermann. We explain how this conjecture would imply as special cases answers to longstanding open questions of Johnson and Lindenstrauss and Kalton.

math.MG↗