Searcharxiv⌕ Search

arXiv subjects

Gennady Samorodnitsky

Publications and source records attributed to Gennady Samorodnitsky.

At least 19 recordsLinked to original sources
arXiv2026

How Robust is Scale-Free Structure? Phase Transitions in Preferential Attachment Model under Attack?

Will an attack on a network affect its structure in a major way? In this work we study this question by investigating a dynamic preferential attachment-type graph in which an attacker interferes with the growth mechanism. We show that our model of an attack exhibits a phase transition. If the skill of the attacker is low, the network preserves its scale-free structure inherited from the underlying preferential attachment structure (even though the tail of the degree distribution becomes lighter due to the attack). When the skill of the attacker increases and crosses a critical boundary, the scale-free property is destroyed. The skill of the attacker is expressed as the probability that, at a given step, the attacker succeeds in preventing adding an edge to the network and, instead, causes a deletion (in a preferential manner) of an edge. In all cases we describe the tail behaviour of the degree distribution.

math.PR↗
arXiv2026

Estimating quantile treatments without strict overlap

We consider the problem of estimating quantile treatment effects without assuming strict overlap, i.e., we do not assume that the propensity score is bounded away from zero. More specifically, we consider an inverse probability weighting (IPW) approach for estimating quantiles in the potential outcomes framework and pay special attention to scenarios where the propensity scores can tend to zero as a regularly varying function. Our approach effectively considers a heavy-tailed objective function for estimating the quantile process. We introduce a truncated IPW estimator that is shown to outperform the standard quantile IPW estimator when strict overlap does not hold. We show that the limiting distribution of the estimated quantile process follows an infinitely divisible law and converges at the rate $n^{1-1/γ}$, where $γ>1$ is the tail index of the propensity scores when they tend to zero. We propose a practical, data-driven procedure for selecting the truncation parameter, grounded in our asymptotic theory. The performance of our estimators is illustrated in numerical experiments and in a dataset that exhibits the presence of extreme propensity scores.

math.ST↗
arXiv2026

Clustering of large deviations in heavy-tailed moving averages: the catastrophe principle in the long-memory case

Clustering of large deviations events in a stationary stochastic process depends critically on the interplay between the tail behavior of the marginal distribution and the strength of temporal dependence. In the class of doubly infinite moving average processes, when the memory is short, previous work has established a sharp contrast in the clustering patterns between light- and heavy-tailed settings, governed by \textit{conspiracy} and \textit{catastrophe} principles respectively. It has also been described how long memory interacts with the conspiracy principle to affect clustering in the light-tailed case. This paper addresses the interaction of long memory with the catastrophe principle in the heavy-tailed case. It turns out that long memory generally allows for a wider range of catastrophes to play a role and leads to longer and qualitatively different clustering patterns.

math.PR↗
arXiv2026

Interlacing sequences resulting from an interval split-merge dynamics and the induced probability measures

We study sequences of partitions of the unit interval into subintervals, starting from the trivial partition, in which each partition is obtained from the one before by splitting its subintervals in two, according to a given rule, and then merging pairs of subintervals at the break points of the old partition. The $n$th partition then comprises $n+1$ subintervals with $n$ break points, which inherently possess an interlacing property. The empirical distribution of these points reveals a surprisingly rich structure, even when the splitting rule is completely deterministic. We consider both deterministic and randomized splitting rules and we study from multiple angles the limiting behavior of the empirical distribution of the break points.

math.PR↗
arXiv2025

Clustering of large deviations events in heavy-tailed moving average processes: the catastrophe principle in the short-memory case

How do large deviation events in a stationary process cluster? The answer depends not only on the type of large deviations, but also on the length of memory in the process. Somewhat unexpectedly, it may also depend on the tails of the process. In this paper we work in the context of large deviations for partial sums in moving average processes with short memory and regularly varying tails. We show that the structure of the large deviation cluster in this case markedly differs from the corresponding structure in the case of exponentially light tails, considered in Chakrabarty and Samorodnitsky (2024). This is due to the difference between the ``conspiracy'' vs. the ``catastrophe'' principles underlying the large deviation events in the light tailed case and the heavy tailed case, correspondingly.

math.PR↗
arXiv2025

Likelihood Inference for Possibly Non-Stationary Processes via Adaptive Overdifferencing

We make an observation that facilitates exact likelihood-based inference for the parameters of the popular ARFIMA model without requiring stationarity by allowing the upper bound $\bar{d}$ for the memory parameter $d$ to exceed $0.5$: estimating the parameters of a single non-stationary ARFIMA model is equivalent to estimating the parameters of a sequence of stationary ARFIMA models. This allows for the use of existing methods for evaluating the likelihood for an invertible and stationary ARFIMA model. This enables improved inference because many standard methods perform poorly when estimates are close to the boundary of the parameter space. It also allows us to leverage the wealth of likelihood approximations that have been introduced for estimating the parameters of a stationary process. We explore how estimation of the memory parameter $d$ depends on the upper bound $\bar{d}$ and introduce adaptive procedures for choosing $\bar{d}$. We show via simulation how our adaptive procedures estimate the memory parameter well, relative to existing alternatives, when the true value is as large as 2.5.

stat.ME↗
arXiv2024

Degree counts in random simplicial complexes of the preferential attachment type

We extend the classical preferential attachment random graph model to random simplicial complexes. At each stage of the model, we choose one of the existing $k$-simplices with probability proportional to its $k$-degree. The chosen $k$-simplex then forms a $(k+1)$-simplex with a newly arriving vertex. We establish a strong law of large numbers for the degree counts across multiple dimensions. The limiting probability mass function is expressed as a mixture of mass functions of different types of negative binomial random variables. This limiting distribution has power-law characteristics and we explore the limiting extremal dependence of the degree counts across different dimensions in the framework of multivariate regular variation. Finally, we prove multivariate weak convergence, under appropriate normalization, of degree counts in different dimensions, of ordered $k$-simplices. The resulting weak limit can be represented as a function of independent linear birth processes with immigration.

math.PR↗
arXiv2024

The Asymptotics of the Expected Betti Numbers of Preferential Attachment Clique Complexes

The preferential attachment model is a natural and popular random graph model for a growing network that contains very well-connected ``hubs''. We study the higher-order connectivity of such a network by investigating the topological properties of its clique complex. We concentrate on the expected Betti numbers, a sequence of topological invariants of the complex related to the numbers of holes of different dimensions. We determine the asymptotic growth rates of the expected Betti numbers, and prove that the expected Betti number at dimension 1 grows linearly fast, while those at higher dimensions grow sublinearly fast. Our theoretical results are illustrated by simulations. (Changes are made in this version to generalize Proposition 14 and to streamline proofs. These changes are shown in blue.)

math.PR↗
arXiv2024

Detection of Small Holes by the Scale-Invariant Robust Density-Aware Distance (RDAD) Filtration

A novel topological-data-analytical (TDA) method is proposed to distinguish, from noise, small holes surrounded by high-density regions of a probability density function. The proposed method is robust against additive noise and outliers. Traditional TDA tools, like those based on the distance filtration, often struggle to distinguish small features from noise, because both have short persistences. An alternative filtration, called the Robust Density-Aware Distance (RDAD) filtration, is proposed to prolong the persistences of small holes of high-density regions. This is achieved by weighting the distance function by the density in the sense of Bell et al. The concept of distance-to-measure is incorporated to enhance stability and mitigate noise. The persistence-prolonging property and robustness of the proposed filtration are rigorously established, and numerical experiments are presented to demonstrate the proposed filtration's utility in identifying small holes.

math.ST↗
arXiv2024

Epsilon*: Privacy Metric for Machine Learning Models

We introduce Epsilon*, a new privacy metric for measuring the privacy risk of a single model instance prior to, during, or after deployment of privacy mitigation strategies. The metric requires only black-box access to model predictions, does not require training data re-sampling or model re-training, and can be used to measure the privacy risk of models not trained with differential privacy. Epsilon* is a function of true positive and false positive rates in a hypothesis test used by an adversary in a membership inference attack. We distinguish between quantifying the privacy loss of a trained model instance, which we refer to as empirical privacy, and quantifying the privacy loss of the training mechanism which produces this model instance. Existing approaches in the privacy auditing literature provide lower bounds for the latter, while our metric provides an empirical lower bound for the former by relying on an ($ε$, $δ$)-type of quantification of the privacy of the trained model instance. We establish a relationship between these lower bounds and show how to implement Epsilon* to avoid numerical and noise amplification instability. We further show in experiments on benchmark public data sets that Epsilon* is sensitive to privacy risk mitigation by training with differential privacy (DP), where the value of Epsilon* is reduced by up to 800% compared to the Epsilon* values of non-DP trained baseline models. This metric allows privacy auditors to be independent of model owners, and enables visualizing the privacy-utility landscape to make informed decisions regarding the trade-offs between model privacy and utility.

cs.LG↗
arXiv2023

Adaptive Privacy Composition for Accuracy-first Mechanisms

In many practical applications of differential privacy, practitioners seek to provide the best privacy guarantees subject to a target level of accuracy. A recent line of work by Ligett et al. '17 and Whitehouse et al. '22 has developed such accuracy-first mechanisms by leveraging the idea of noise reduction that adds correlated noise to the sufficient statistic in a private computation and produces a sequence of increasingly accurate answers. A major advantage of noise reduction mechanisms is that the analysts only pay the privacy cost of the least noisy or most accurate answer released. Despite this appealing property in isolation, there has not been a systematic study on how to use them in conjunction with other differentially private mechanisms. A fundamental challenge is that the privacy guarantee for noise reduction mechanisms is (necessarily) formulated as ex-post privacy that bounds the privacy loss as a function of the released outcome. Furthermore, there has yet to be any study on how ex-post private mechanisms compose, which allows us to track the accumulated privacy over several mechanisms. We develop privacy filters [Rogers et al. '16, Feldman and Zrnic '21, and Whitehouse et al. '22'] that allow an analyst to adaptively switch between differentially private and ex-post private mechanisms subject to an overall differential privacy guarantee.

cs.CR↗
arXiv2023

Empirical Risk Minimization for Losses without Variance

This paper considers an empirical risk minimization problem under heavy-tailed settings, where data does not have finite variance, but only has $p$-th moment with $p \in (1,2)$. Instead of using estimation procedure based on truncated observed data, we choose the optimizer by minimizing the risk value. Those risk values can be robustly estimated via using the remarkable Catoni's method (Catoni, 2012). Thanks to the structure of Catoni-type influence functions, we are able to establish excess risk upper bounds via using generalized generic chaining methods. Moreover, we take computational issues into consideration. We especially theoretically investigate two types of optimization methods, robust gradient descent algorithm and empirical risk-based methods. With an extensive numerical study, we find that the optimizer based on empirical risks via Catoni-style estimation indeed shows better performance than other baselines. It indicates that estimation directly based on truncated data may lead to unsatisfactory results.

stat.ML↗
arXiv2023

Limit theorems for high-dimensional Betti numbers in the multiparameter random simplicial complexes

We consider the multiparameter random simplicial complex on a vertex set $\{ 1,\dots,n \}$, which is parameterized by multiple connectivity probabilities. Our key results concern the topology of this complex of dimensions higher than the critical dimension. We show that the higher-dimensional Betti numbers satisfy strong laws of large numbers and central limit theorems. Moreover, lower tail large deviations for these Betti numbers are also discussed. Some of our results indicate an occurrence of phase transitions in terms of the scaling constants of the central limit theorem, and the exponentially decaying rate of convergence of lower tail large deviation probabilities.

math.PR↗
arXiv2023

A Cover Time Study of a non-Markovian Algorithm

Given a traversal algorithm, cover time is the expected number of steps needed to visit all nodes in a given graph. A smaller cover time means a higher exploration efficiency of traversal algorithm. Although random walk algorithms have been studied extensively in the existing literature, there has been no cover time result for any non-Markovian method. In this work, we stand on a theoretical perspective and show that the negative feedback strategy (a count-based exploration method) is better than the naive random walk search. In particular, the former strategy can locally improve the search efficiency for an arbitrary graph. It also achieves smaller cover times for special but important graphs, including clique graphs, tree graphs, etc. Moreover, we make connections between our results and reinforcement learning literature to give new insights on why classical UCB and MCTS algorithms are so useful. Various numerical results corroborate our theoretical findings.

cs.DS↗
arXiv2023

Spectral learning of multivariate extremes

We propose a spectral clustering algorithm for analyzing the dependence structure of multivariate extremes. More specifically, we focus on the asymptotic dependence of multivariate extremes characterized by the angular or spectral measure in extreme value theory. Our work studies the theoretical performance of spectral clustering based on a random $k$-nearest neighbor graph constructed from an extremal sample, i.e., the angular part of random vectors for which the radius exceeds a large threshold. In particular, we derive the asymptotic distribution of extremes arising from a linear factor model and prove that, under certain conditions, spectral clustering can consistently identify the clusters of extremes arising in this model. Leveraging this result we propose a simple consistent estimation strategy for learning the angular measure. Our theoretical findings are complemented with numerical experiments illustrating the finite sample performance of our methods.

stat.ML↗
arXiv2023

Clustering of large deviations in moving average processes: the long memory regime

We investigate how large deviations events cluster in the framework of an infinite moving average process with light-tailed noise and long memory. The long memory makes clusters larger, and the asymptotic behaviour of the size of the cluster turns out to be described by the first hitting time of a randomly shifted fractional Brownian motion with drift.

math.PR↗
arXiv2022

Kernel PCA for multivariate extremes

We propose kernel PCA as a method for analyzing the dependence structure of multivariate extremes and demonstrate that it can be a powerful tool for clustering and dimension reduction. Our work provides some theoretical insight into the preimages obtained by kernel PCA, demonstrating that under certain conditions they can effectively identify clusters in the data. We build on these new insights to characterize rigorously the performance of kernel PCA based on an extremal sample, i.e., the angular part of random vectors for which the radius exceeds a large threshold. More specifically, we focus on the asymptotic dependence of multivariate extremes characterized by the angular or spectral measure in extreme value theory and provide a careful analysis in the case where the extremes are generated from a linear factor model. We give theoretical guarantees on the performance of kernel PCA preimages of such extremes by leveraging their asymptotic distribution together with Davis-Kahan perturbation bounds. Our theoretical findings are complemented with numerical experiments illustrating the finite sample performance of our methods.

stat.ML↗