Searcharxiv⌕ Search

arXiv subjects

Roberto Imbuzeiro Oliveira

Publications and source records attributed to Roberto Imbuzeiro Oliveira.

At least 19 recordsLinked to original sources
arXiv2026

Debiasing the Lasso under Weaker Tail Assumptions

We consider the problem of high-dimensional inference with the lasso estimator. Different methods including 'double selection' techniques and multiple versions of the 'debiased lasso' have been proposed for this task with noticeable success. However, most guarantees assume strong hypotheses on the underlying data process and the errors in the linear regression model, such as subgaussian designs and independence between errors and the data itself. We show that 'standardizing' one's dataset -- a natural procedure in practical penalized regression -- leads to the same results under much weaker hypotheses, paying only a small price for not assuming light tails. The key technical point allowed by this step is exploiting the concentration properties of self-normalized processes. Importantly, we prove our results for two different methods closely related to the 'debiased lasso'. The second method performs valid inference even for a misspecified linear model, under mild sparsity conditions similar to the 'double selection' literature.

math.ST↗
arXiv2026

Nonparametric inference on Fokker-Plank and McKean-Vlasov models

We propose a kernel-based estimator of the velocity field governing the transport and diffusion of $d$-dimensional interacting particles. Assuming the initial positions are i.i.d. with law $μ_0$, we establish consistency of the estimator with an explicit mean-squared error rate of order $h^2 + N^(-2/(d+2))$, where $h$ denotes the time-discretization step and N the number of particles. The analysis covers two distinct settings: the Fokker-Planck equation, where we recover the underlying potential function, and the McKean-Vlasov equation, where we deconvolve the interaction kernel driving the mean-field dynamics.

math.ST↗
arXiv2026

Geometric planted matchings beyond the Gaussian model

We consider the problem of recovering an unknown matching between a set of $n$ randomly placed points in $\mathbb{R}^d$ and random perturbations of these points. This can be seen as a model for particle tracking and more generally, entity resolution. We use matchings in random geometric graphs to derive minimax lower bounds for this problem that hold under great generality. Using these results we show that for a fixed $d$, as long as the noise distribution has finite $d$-th moment, and both initial positions and noise have bounded continuous densities, the minimax rate for the problem scales as $Θ(n^2σ^d \wedge n)$. Under the stronger assumptions that the tail of the noise is sub-Gaussian, we show that the order of the number of mistakes made by an estimator that minimizes the sum of squared Euclidean distances is minimax optimal when $d$ is fixed and is optimal up to $n^{o(1)}$ factors when $d = o(\log n)$. In the high-dimensional regime we consider a setup where both initial positions and perturbations have independent sub-Gaussian coordinates. In this setup we give sufficient conditions under which the same estimator makes no mistakes with high probability. We prove an analogous result for an adapted version of this estimator that incorporates information on the covariance matrix of the perturbations.

math.ST↗
arXiv2026

Mean Testing under Truncation beyond Gaussian

We characterize the fundamental limits of high-dimensional mean testing under arbitrary truncation, where samples are drawn from the conditional distribution $P(\cdot \mid S)$ for an unknown truncation set $S$ that may hide up to an $\varepsilon$-fraction of the probability mass. For distributions with $p$-th directional moments of magnitude at most $ν_{P,p}$, truncation induces a bias of order $O(ν_{P,p}\varepsilon^{1-1/p})$. This bias creates a sharp information-theoretic detectability floor: when the signal $α$ falls below this threshold, the null and alternative hypotheses are indistinguishable even with infinite data. Above this floor, we prove that a simple second-order test achieving near-optimal sample complexity $n = O\!\left(\frac{\|Σ_P\|}{(α-4ν_{P,p}\varepsilon^{1-1/p})^2}\sqrt{d}\right)$. We further identify a structural escape from this finite-moment bias barrier. Under a directional median regularity assumption, truncation bias improves to linear order $O(\varepsilon)$. This reveals an intermediate regime in which estimation requires $Θ(d)$ samples for uniform recovery, while testing recovers the classical $Θ(\sqrt d)$ rate once truncation bias is eliminated. Together, our results provide a unified framework for mean testing under truncation, connecting finite-moment, sub-Gaussian, and median-regular structural regimes.

stat.ML↗
arXiv2025

Robust, sub-Gaussian mean estimators in metric spaces

Estimating the mean of a random vector from i.i.d. data has received considerable attention, and the optimal accuracy one may achieve with a given confidence is fairly well understood by now. When the data take values in more general metric spaces, an appropriate extension of the notion of the mean is the Fréchet mean. While asymptotic properties of the most natural Fréchet mean estimator (the empirical Fréchet mean) have been thoroughly researched, non-asymptotic performance bounds have only been studied recently. The aim of this paper is to study the performance of estimators of the Fréchet mean in general metric spaces under possibly heavy-tailed and contaminated data. In such cases, the empirical Fréchet mean is a poor estimator. We propose a general estimator based on high-dimensional extensions of trimmed means and prove general performance bounds. Unlike all previously established bounds, ours generalize the optimal bounds known for Euclidean data. The main message of the bounds is that, much like in the Euclidean case, the optimal accuracy is governed by two "variance" terms: a "global variance" term that is independent of the prescribed confidence, and a potentially much smaller, confidence-dependent "local variance" term. We apply our results for metric spaces with curvature bounded from below, such as Wasserstein spaces, and for uniformly convex Banach spaces.

math.ST↗
arXiv2025

On the quality of randomized approximations of Tukey's depth

Tukey's depth (or halfspace depth) is a widely used measure of centrality for multivariate data. However, exact computation of Tukey's depth is known to be a hard problem in high dimensions. As a remedy, randomized approximations of Tukey's depth have been proposed. In this paper we explore when such randomized algorithms return a good approximation of Tukey's depth. We study the case when the data are sampled from a log-concave isotropic distribution. We prove that, if one requires that the algorithm runs in polynomial time in the dimension, the randomized algorithm correctly approximates the maximal depth $1/2$ and depths close to zero. On the other hand, for any point of intermediate depth, any good approximation requires exponential complexity.

stat.ML↗
arXiv2024

The top eigenvalue of uniformly random trees

Let ${\mathbf T}_n$ be a uniformly random tree with vertex set $[n]=\{1,\ldots,n\}$, let $Δ_{{\mathbf T}_n}$ be the largest vertex degree in ${\mathbf T}_n$, and let $λ_1({\mathbf T}_n),\ldots,λ_n({\mathbf T}_n)$ be the eigenvalues of its adjacency matrix, arranged in decreasing order. We prove that $|λ_1({\mathbf T}_n)-\sqrt{Δ_{{\mathbf T}_n}}| \to 0$ in expectation as $n \to \infty$, and additionally prove probability tail bounds for $|λ_1({\mathbf T}_n)-\sqrt{Δ_{{\mathbf T}_n}}|$. Writing $a_n$ for any median of $Δ_{{\mathbf T}_n}$, we also prove that $|λ_k({\mathbf T}_n)-\sqrt{a_n}| \to 0$ in expectation, uniformly over $1 \le k \le e^{\log^β(n)}$, for any fixed $β\in (0,1/2)$. The proof is based on the trace method and thus on counting closed walks in a random tree. To this end, we develop novel combinatorial tools for encoding walks in trees that we expect will find other applications. In order to apply these tools, we show that uniformly random trees -- after appropriate "surgery" -- satisfy, with high probability, the properties required for the combinatorial bounds to be effective.

math.PR↗
arXiv2024

Subtractive random forests

Motivated by online recommendation systems, we study a family of random forests. The vertices of the forest are labeled by integers. Each non-positive integer $i\le 0$ is the root of a tree. Vertices labeled by positive integers $n \ge 1$ are attached sequentially such that the parent of vertex $n$ is $n-Z_n$, where the $Z_n$ are i.i.d.\ random variables taking values in $\mathbb N$. We study several characteristics of the resulting random forest. In particular, we establish bounds for the expected tree sizes, the number of trees in the forest, the number of leaves, the maximum degree, and the height of the forest. We show that for all distributions of the $Z_n$, the forest contains at most one infinite tree, almost surely. If ${\mathbb E} Z_n < \infty$, then there is a unique infinite tree and the total size of the remaining trees is finite, with finite expected value if ${\mathbb E}Z_n^2 < \infty$. If ${\mathbb E} Z_n = \infty$ then almost surely all trees are finite.

math.PR↗
arXiv2021

The contact process over a dynamical d-regular graph

We consider the contact process on a dynamic graph defined as a random $d$-regular graph with a stationary edge-switching dynamics. In this graph dynamics, independently of the contact process state, each pair $\{e_1,e_2\}$ of edges of the graph is replaced by new edges $\{e_1',e_2'\}$ in a crossing fashion: each of $e_1',e_2'$ contains one vertex of $e_1$ and one vertex of $e_2$. As the number of vertices of the graph is taken to infinity, we scale the rate of switching in a way that any fixed edge is involved in a switching with a rate that approaches a limiting value $\mathsf{v}$, so that locally the switching is seen in the same time scale as that of the contact process. We prove that if the infection rate of the contact process is above a threshold value $\barλ$ (depending on $d$ and $\mathsf{v}$), then the infection survives for a time that grows exponentially with the size of the graph. By proving that $\barλ$ is strictly smaller than the lower critical infection rate of the contact process on the infinite $d$-regular tree, we show that there are values of $λ$ for which the infection dies out in logarithmic time in the static graph but survives exponentially long in the dynamic graph.

math.PR↗
arXiv2016

Stochastic processes with random contexts: a characterization, and adaptive estimators for the transition probabilities

This paper introduces the concept of random context representations for the transition probabilities of a finite-alphabet stochastic process. Processes with these representations generalize context tree processes (a.k.a. variable length Markov chains), and are proven to coincide with processes whose transition probabilities are almost surely continuous functions of the (infinite) past. This is similar to a classical result by Kalikow about continuous transition probabilities. Existence and uniqueness of a minimal random context representation are proven, and an estimator of the transition probabilities based on this representation is shown to have very good "pastwise adaptativity" properties. In particular, it achieves minimax performance, up to logarithmic factors, for binary renewal processes with bounded $2+γ$ moments.

math.PR↗
arXiv2015

Exceptional rotations of random graphs: a VC theory

In this paper we explore maximal deviations of large random structures from their typical behavior. We introduce a model for a high-dimensional random graph process and ask analogous questions to those of Vapnik and Chervonenkis for deviations of averages: how "rich" does the process have to be so that one sees atypical behavior. In particular, we study a natural process of Erdős-Rényi random graphs indexed by unit vectors in $\mathbb{R}^d$. We investigate the deviations of the process with respect to three fundamental properties: clique number, chromatic number, and connectivity. In all cases we establish upper and lower bounds for the minimal dimension $d$ that guarantees the existence of "exceptional directions" in which the random graph behaves atypically with respect to the property. For each of the three properties, four theorems are established, to describe upper and lower bounds for the threshold dimension in the subcritical and supercritical regimes.

math.PR↗
arXiv2013

The lower tail of random quadratic forms, with applications to ordinary least squares and restricted eigenvalue properties

Finite sample properties of random covariance-type matrices have been the subject of much research. In this paper we focus on the "lower tail" of such a matrix, and prove that it is subgaussian under a simple fourth moment assumption on the one-dimensional marginals of the random vectors. A similar result holds for more general sums of random positive semidefinite matrices, and the (relatively simple) proof uses a variant of the so-called PAC-Bayesian method for bounding empirical processes. We give two applications of the main result. In the first one we obtain a new finite-sample bound for ordinary least squares estimator in linear regression with random design. Our result is model-free, requires fairly weak moment assumptions and is almost optimal. Our second application is to bounding restricted eigenvalue constants of certain random ensembles with "heavy tails". These constants are important in the analysis of problems in Compressed Sensing and High Dimensional Statistics, where one recovers a sparse vector from a small umber of linear measurements. Our result implies that heavy tails still allow for the fast recovery rates found in efficient methods such as the LASSO and the Dantzig selector. Along the way we strengthen, with a fairly short argument, a recent result of Rudelson and Zhou on the restricted eigenvalue property.

math.PR↗
arXiv2013

Mean field conditions for coalescing random walks

The main results in this paper are about the full coalescence time $\mathsf{C}$ of a system of coalescing random walks over a finite graph $G$. Letting $\mathsf{m}(G)$ denote the mean meeting time of two such walkers, we give sufficient conditions under which $\mathbf{E}[\mathsf{C}]\approx 2\mathsf{m}(G)$ and $\mathsf{C}/\mathsf{m}(G)$ has approximately the same law as in the "mean field" setting of a large complete graph. One of our theorems is that mean field behavior occurs over all vertex-transitive graphs whose mixing times are much smaller than $\mathsf{m}(G)$; this nearly solves an open problem of Aldous and Fill and also generalizes results of Cox for discrete tori in $d\geq2$ dimensions. Other results apply to nonreversible walks and also generalize previous theorems of Durrett and Cooper et al. Slight extensions of these results apply to voter model consensus times, which are related to coalescing random walks via duality. Our main proof ideas are a strengthening of the usual approximation of hitting times by exponential random variables, which give results for nonstationary initial states; and a new general set of conditions under which we can prove that the hitting time of a union of sets behaves like a minimum of independent exponentials. In particular, this will show that the first meeting time among $k$ random walkers has mean $\approx\mathsf{m}(G)/\bigl({\matrix{k 2}}\bigr)$.

math.PR↗
arXiv2013

Mixing of the symmetric exclusion processes in terms of the corresponding single-particle random walk

We prove an upper bound for the $\varepsilon$-mixing time of the symmetric exclusion process on any graph G, with any feasible number of particles. Our estimate is proportional to $\mathsf{T}_{\mathsf{RW}(G)}\ln(|V|/\varepsilon)$, where |V| is the number of vertices in G, and $\mathsf{T}_{\mathsf{RW}(G)}$ is the 1/4-mixing time of the corresponding single-particle random walk. This bound implies new results for symmetric exclusion on expanders, percolation clusters, the giant component of the Erdos-Renyi random graph and Poisson point processes in $\mathbb{R}^d$. Our technical tools include a variant of Morris's chameleon process.

math.PR↗
arXiv2012

Tight inequalities among set hitting times in Markov chains

Given an irreducible discrete-time Markov chain on a finite state space, we consider the largest expected hitting time $T(α)$ of a set of stationary measure at least $α$ for $α\in(0,1)$. We obtain tight inequalities among the values of $T(α)$ for different choices of $α$. One consequence is that $T(α) \le T(1/2)/α$ for all $α< 1/2$. As a corollary we have that, if the chain is lazy in a certain sense as well as reversible, then $T(1/2)$ is equivalent to the chain's mixing time, answering a question of Peres. We furthermore demonstrate that the inequalities we establish give an almost everywhere pointwise limiting characterisation of possible hitting time functions $T(α)$ over the domain $α\in(0,1/2]$.

math.PR↗
arXiv2012

Mixing and hitting times for finite Markov chains

Let 0<α<1/2. We show that the mixing time of a continuous-time reversible Markov chain on a finite state space is about as large as the largest expected hitting time of a subset of stationary measure at least αof the state space. Suitably modified results hold in discrete time and/or without the reversibility assumption. The key technical tool is a construction of a random set A such that the hitting time of A is both light-tailed and a stationary time for the chain. We note that essentially the same results were obtained independently by Peres and Sousi [arXiv:1108.0133].

math.PR↗
arXiv2010

On the coalescence time of reversible random walks

Consider a system of coalescing random walks where each individual performs random walk over a finite graph G, or (more generally) evolves according to some reversible Markov chain generator Q. Let C be the first time at which all walkers have coalesced into a single cluster. C is closely related to the consensus time of the voter model for this G or Q. We prove that the expected value of C is at most a constant multiple of the largest hitting time of an element in the state space. This solves a problem posed by Aldous and Fill and gives sharp bounds in many examples, including all vertex-transitive graphs. We also obtain results on the expected time until only k>1 clusters remain. Our proof tools include a new exponential inequality for the meeting time of a reversible Markov chain and a deterministic trajectory, which we believe to be of independent interest.

math.PR↗