Searcharxiv⌕ Search

arXiv subjects

Jian Ding

Publications and source records attributed to Jian Ding.

At least 145 records · Page 8Linked to original sources
arXiv2016

Liouville first passage percolation: the weight exponent is strictly less than 1 at high temperatures

Let $\{η_{N, v}: v\in V_N\}$ be a discrete Gaussian free field in a two-dimensional box $V_N$ of side length $N$ with Dirichlet boundary conditions. We study the Liouville first passage percolation, i.e., the shortest path metric where each vertex is given a weight of $e^{γη_{N, v}}$ for some $γ>0$. We show that for sufficiently small but fixed $γ>0$, the expected Liouville FPP distance between any pair of vertices is $O(N^{1-γ^2/10^3})$.

math.PR↗
arXiv2015

Testing for high-dimensional geometry in random graphs

We study the problem of detecting the presence of an underlying high-dimensional geometric structure in a random graph. Under the null hypothesis, the observed graph is a realization of an Erdős-Rényi random graph $G(n,p)$. Under the alternative, the graph is generated from the $G(n,p,d)$ model, where each vertex corresponds to a latent independent random vector uniformly distributed on the sphere $\mathbb{S}^{d-1}$, and two vertices are connected if the corresponding latent vectors are close enough. In the dense regime (i.e., $p$ is a constant), we propose a near-optimal and computationally efficient testing procedure based on a new quantity which we call signed triangles. The proof of the detection lower bound is based on a new bound on the total variation distance between a Wishart matrix and an appropriately normalized GOE matrix. In the sparse regime, we make a conjecture for the optimal detection boundary. We conclude the paper with some preliminary steps on the problem of estimating the dimension in $G(n,p,d)$.

math.ST↗
arXiv2015

Percolation of averages in the stochastic mean field model: the near-supercritical regime

For a complete graph of size $n$, assign each edge an i.i.d.\ exponential variable with mean $n$. For $λ>0$, consider the length of the longest path whose average weight is at most $λ$. It was shown by Aldous (1998) that the length is of order $\log n$ for $λ< 1/\mathrm{e}$ and of order $n$ for $λ> 1/\mathrm{e}$. In this paper, we study the near-supercritical regime where $λ= \mathrm{e}^{-1} +η$ with $η>0$ a small fixed number. We show that there exist two absolute constants $c^*, C^*>0$ such that with high probability the length is in between $n \mathrm{e}^{-C^*/\sqrtη}$ and $n \mathrm{e}^{-c^*/\sqrtη}$. Our result corrects a non-rigorous prediction of Aldous (2005).

math.PR↗
arXiv2015

Distances in critical long range percolation

We study the long range percolation model on $\mathbb{Z}$ where sites $i$ and $j$ are connected with probability $β|i-j|^{-s}$. Graph distances are now well understood for all exponents $s$ except in the case $s=2$ where the model exhibits non-trivial self-similar scaling. Establishing a conjecture of Benjamini and Berger \cite{BenBer:01}, we prove that the typical distance from site 0 to $n$ grows as a power law $n^{θ(β)}$ up to a multiplicative constant for some exponent $0<θ(β)<1$ as does the diameter of the graph on a box of length $n$.

math.PR↗
arXiv2015

Supercritical minimum mean-weight cycles

We study the weight and length of the minimum mean-weight cycle in the stochastic mean-field distance model, i.e., in the complete graph on $n$ vertices with edges weighted by independent exponential random variables. Mathieu and Wilson showed that the minimum mean-weight cycle exhibits one of two distinct behaviors, according to whether its mean weight is smaller or larger than $1/(ne)$; and that both scenarios occur with positive probability in the limit $n\to\infty$. If the mean weight is $< 1/(ne)$, the length is of constant order. If the mean weight is $> 1/(ne)$, it is concentrated just above $1/(n e)$, and the length diverges with $n$. The analysis of Mathieu--Wilson gives a detailed characterization of the subcritical regime, including the (non-degenerate) limiting distributions of the weight and length, but leaves open the supercritical behavior. We determine the asymptotics for the supercritical regime, showing that with high probability, the minimum mean weight is $(n e)^{-1}[1 + π^2/(2 \log^2 n) + O((\log n)^{-3})]$, and the cycle achieving this minimum has length on the order of $(\log n)^3$.

math.PR↗
arXiv2015

Convergence of the centered maximum of log-correlated Gaussian fields

We show that the centered maximum of a sequence of log-correlated Gaussian fields in any dimension converges in distribution, under the assumption that the covariances of the fields converge in a suitable sense. We identify the limit as a randomly shifted Gumbel distribution, and characterize the random shift as the limit in distribution of a sequence of random variables, reminiscent of the derivative martingale in the theory of Branching Random Walk and Gaussian Chaos. We also discuss applications of the main convergence theorem and discuss examples that show that for logarithmically correlated fields, some additional structural assumptions of the type we make are needed for convergence of the centered maximum.

math.PR↗
arXiv2014

Online Learning with Composite Loss Functions

We study a new class of online learning problems where each of the online algorithm's actions is assigned an adversarial value, and the loss of the algorithm at each step is a known and deterministic function of the values assigned to its recent actions. This class includes problems where the algorithm's loss is the minimum over the recent adversarial values, the maximum over the recent values, or a linear combination of the recent values. We analyze the minimax regret of this class of problems when the algorithm receives bandit feedback, and prove that when the minimum or maximum functions are used, the minimax regret is $\tilde Ω(T^{2/3})$ (so called hard online learning problems), and when a linear function is used, the minimax regret is $\tilde O(\sqrt{T})$ (so called easy learning problems). Previously, the only online learning problem that was known to be provably hard was the multi-armed bandit with switching costs.

cs.LG↗
arXiv2014

Asymptotics of cover times via Gaussian free fields: Bounded-degree graphs and general trees

In this paper we show that on bounded degree graphs and general trees, the cover time of the simple random walk is asymptotically equal to the product of the number of edges and the square of the expected supremum of the Gaussian free field on the graph, assuming that the maximal hitting time is significantly smaller than the cover time. Previously, this was only proved for regular trees and the 2D lattice. Furthermore, for general trees, we derive exponential concentration for the cover time, which implies that the standard deviation of the cover time is bounded by the geometric mean of the cover time and the maximal hitting time.

math.PR↗
arXiv2013

Mixing under monotone censoring

We initiate the study of mixing times of Markov chain under monotone censoring. Suppose we have some Markov Chain $M$ on a state space $Ω$ with stationary distribution $π$ and a monotone set $A \subset Ω$. We consider the chain $M'$ which is the same as the chain $M$ started at some $x \in A$ except that moves of $M$ of the form $x \to y$ where $x \in A$ and $y \notin A$ are {\em censored} and replaced by the move $x \to x$. If $M$ is ergodic and $A$ is connected, the new chain converges to $π$ conditional on $A$. In this paper we are interested in the mixing time of the chain $M'$ in terms of properties of $M$ and $A$. Our results are based on new connections with the field of property testing. A number of open problems are presented.

math.PR↗
arXiv2013

On multiple peaks and moderate deviations for supremum of Gaussian field

We prove two theorems concerning extreme values of general Gaussian fields. Our first theorem concerns with the concept of multiple peaks. A theorem of Chatterjee states that when a centered Gaussian field admits the so-called superconcentration property, it typically attains values near its maximum on multiple near-orthogonal sites, known as multiple peaks. We improve his theorem in two aspects: (i) the number of peaks attained by our bound is of the order $\exp(c / σ^2)$ (as opposed to Chatterjee's polynomial bound in $1/σ$), where $σ$ is the standard deviation of the supremum of the Gaussian field, which is assumed to have variance at most $1$ and (ii) our bound need not assume that the correlations are non-negative. We also prove a similar result based on the superconcentration of the free energy. As primary applications, we infer that for the S-K spin glass model on the $n$-hypercube and directed polymers on $\mathbb{Z}_n^2$, there are polynomially (in $n$) many near-orthogonal sites that achieve values near their respective maxima. Our second theorem gives an upper bound on moderate deviation for the supremum of a general Gaussian field. While the Gaussian isoperimetric inequality implies a sub-Gaussian concentration bound for the supremum, we show that the exponent in that bound can be improved under the assumption that the expectation of the supremum is of the same order as that of the independent case.

math.PR↗
arXiv2013

Bandits with Switching Costs: T^{2/3} Regret

We study the adversarial multi-armed bandit problem in a setting where the player incurs a unit cost each time he switches actions. We prove that the player's $T$-round minimax regret in this setting is $\widetildeΘ(T^{2/3})$, thereby closing a fundamental gap in our understanding of learning with bandit feedback. In the corresponding full-information version of the problem, the minimax regret is known to grow at a much slower rate of $Θ(\sqrt{T})$. The difference between these two rates provides the \emph{first} indication that learning with bandit feedback can be significantly harder than learning with full-information feedback (previous results only showed a different dependence on the number of actions, but not on $T$.) In addition to characterizing the inherent difficulty of the multi-armed bandit problem with switching costs, our results also resolve several other open problems in online learning. One direct implication is that learning with bandit feedback against bounded-memory adaptive adversaries has a minimax regret of $\widetildeΘ(T^{2/3})$. Another implication is that the minimax regret of online learning in adversarial Markov decision processes (MDPs) is $\widetildeΘ(T^{2/3})$. The key to all of our results is a new randomized construction of a multi-scale random walk, which is of independent interest and likely to prove useful in additional settings.

cs.LG↗
arXiv2013

On level sets of Gaussian fields

In this short note, we present a theorem concerning certain "additive structure" for the level sets of non-degenerate Gaussian fields, which yields the multiple valley phenomenon for extremal fields with exponentially many valleys.

math.PR↗
arXiv2013

Satisfiability threshold for random regular NAE-SAT

We consider the random regular $k$-NAE-SAT problem with $n$ variables each appearing in exactly $d$ clauses. For all $k$ exceeding an absolute constant $k_0$, we establish explicitly the satisfiability threshold $d_*=d_*(k)$. We prove that for $d d_*$ the problem is unsatisfiable with high probability. If the threshold $d_*$ lands exactly on an integer, we show that the problem is satisfiable with probability bounded away from both zero and one. This is the first result to locate the exact satisfiability threshold in a random constraint satisfaction problem exhibiting the condensation phenomenon identified by Krzakala et al. (2007). Our proof verifies the one-step replica symmetry breaking formalism for this model. We expect our methods to be applicable to a broad range of random constraint satisfaction problems and combinatorial problems on random graphs.

math.PR↗
arXiv2013

Maximum independent sets on random regular graphs

We determine the asymptotics of the independence number of the random $d$-regular graph for all $d \ge d_0$. It is highly concentrated, with constant-order fluctuations around $nα_* - c_*\log n$ for explicit constants $α_*(d)$ and $c_*(d)$. Our proof rigorously confirms the one-step replica symmetry breaking heuristics for this problem, and we believe the techniques will be more broadly applicable to the study of other combinatorial properties of random graphs.

math.PR↗
arXiv2013

Mixing time for the Ising model: a uniform lower bound for all graphs

Consider Glauber dynamics for the Ising model on a graph of $n$ vertices. Hayes and Sinclair showed that the mixing time for this dynamics is at least $n\log n/f(Δ)$, where $Δ$ is the maximum degree and $f(Δ) = Θ(Δ\log^2 Δ)$. Their result applies to more general spin systems, and in that generality, they showed that some dependence on $Δ$ is necessary. In this paper, we focus on the ferromagnetic Ising model and prove that the mixing time of Glauber dynamics on any $n$-vertex graph is at least $(1/4+o(1))n \log n$.

math.PR↗
arXiv2013

Markov type and threshold embeddings

For two metric spaces X and Y, say that X {threshold-embeds} into Y if there exist a number K > 0 and a family of Lipschitz maps $f_τ : X \to Y : τ> 0 \}$ such that for every $x,y \in X$, \[ d_X(x,y) \geq τ=> d_Y(f_τ(x),f_τ(y)) \geq \|φ_τ\|_{\Lip} τ/K \] where $\|f_τ\|_{\Lip}$ denotes the Lipschitz constant of $f_τ$. We show that if a metric space X threshold-embeds into a Hilbert space, then X has Markov type 2. As a consequence, planar graph metrics and doubling metrics have Markov type 2, answering questions of Naor, Peres, Schramm, and Sheffield. More generally, if a metric space X threshold-embeds into a p-uniformly smooth Banach space, then X has Markov type p. This suggests some non-linear analogs of Kwapien's theorem. For instance, a subset $X \subseteq L_1$ threshold-embeds into Hilbert space if and only if X has Markov type 2.

math.MG↗
arXiv2013

Extreme values for two-dimensional discrete Gaussian free field

We consider in this paper the collection of near maxima of the discrete, two dimensional Gaussian free field in a box with Dirichlet boundary conditions. We provide a rough description of the geometry of the set of near maxima, estimates on the gap between the two largest maxima, and an estimate for the right tail up to a multiplicative constant on the law of the centered maximum.

math.PR↗