Searcharxiv⌕ Search

arXiv subjects

Olgica Milenkovic

Publications and source records attributed to Olgica Milenkovic.

At least 19 recordsLinked to original sources
arXiv2026

WaveletDiff: Multilevel Wavelet Diffusion For Time Series Generation

Time series are ubiquitous in many applications that involve forecasting, classification and causal inference tasks, such as healthcare, finance, audio signal processing and climate sciences. Still, large, high-quality time series datasets remain scarce. Synthetic generation can address this limitation; however, current models confined either to the time or frequency domains struggle to reproduce the inherently multi-scaled structure of real-world time series. We introduce WaveletDiff, a new framework that trains diffusion models directly on wavelet coefficients to exploit the inherent multi-resolution structure of time series data. The model combines dedicated transformers for each decomposition level with cross-level attention mechanisms that enable selective information exchange between temporal and frequency scales through adaptive gating. It is also informed by level-specific energy constraints based on Parseval's theorem which preserve time-frequency properties throughout the diffusion process. Comprehensive tests across six real-world datasets from energy, finance, and neuroscience domains demonstrate that WaveletDiff outperforms the diffusion baselines FourierDiffusion, Diffusion-TS, and SigDiffusions on the majority of metrics, with the smallest margin over FourierDiffusion, while still achieving roughly 3x lower discriminative and Context-FID scores. Against the VAE/transformer-based MSDformer, the results are mostly comparable, with WaveletDiff using fewer parameters and less training time on most datasets. The most revealing finding is the significant performance gap on fMRI data (in favor of MSDformer) and EEG (in favor of WaveletDiff). This finding is explained via a careful testing/examination of the properties of wavelet coefficients for generative, as opposed to analyses/decomposition tasks. Our code is available at https://github.com/GarlicWang/WaveletDiff.

cs.LG↗
arXiv2026

Positional Identifiability from Pairwise Collision Data

We study the problem of recovering the relative positions of objects moving along the real line based only on pairwise collision data. While interaction-based sensing systems arise naturally in a variety of practical settings, a systematic theoretical understanding of positional identifiability from collision observations alone remains unexplored. Our contributions are three-fold. First, under the full observability model, in which both the set of collisions and their temporal ordering are known, we show that the relative positions of all objects can be uniquely recovered if and only if the collision history, represented as a graph, is connected. Second, we show that under partial observability, where only colliding pairs are observed without timing information, the problem is related to \emph{function graphs} and introduce a canonical layer decomposition in which each layer corresponds to a maximal clique; the contraction graph induced by this decomposition is an interval graph, and we provide efficient algorithms to recover it. Third, under incomplete observations where even some pairwise collision observations may be missing, we formulate the problem as a graph completion problem and establish its NP-hardness via a $4$-approximation relationship with the graph bandwidth problem.

cs.DM↗
arXiv2026

Correcting Contextual Deletions in DNA Nanopore Readouts

The problem of designing codes for deletion-correction and synchronization has received renewed interest due to applications in DNA-based data storage systems that use nanopore sequencers as readout platforms. In almost all instances, deletions are assumed to be imposed independently of each other and of the sequence context. These assumptions are not valid in practice, since nanopore errors tend to occur within specific contexts. We study contextual nanopore deletion-errors through the example setting of deterministic single deletions following (complete) runlengths of length at least $k$. The model critically depends on the runlength threshold $k$, and we examine two regimes for $k$: a) $k=C\log n$ for a constant $C\in(0,1)$; in this case, we study error-correcting codes that can protect from a constant number $t$ of contextual deletions, and show that the minimum redundancy (ignoring lower-order terms) is between $(1-C)t\log n$ and $2(1-C)t\log n$, meaning that it is a ($1-C$)-fraction of that of arbitrary $t$-deletion-correcting codes. To complement our non-constructive redundancy upper bound, we design efficiently and encodable and decodable codes for any constant $t$. In particular, for $t=1$ and $C>1/2$ we construct efficient codes with redundancy that essentially matches our non-constructive upper bound; b) $k$ equal a constant; in this case we consider the extremal problem where the number of deletions is not bounded and a deletion is imposed after every run of length at least $k$, which we call the extremal contextual deletion channel. This combinatorial setting arises naturally by considering a probabilistic channel that introduces contextual deletions after each run of length at least $k$ with probability $p$ and taking the limit $p\to 1$. We obtain sharp bounds on the maximum achievable rate under the extremal contextual deletion channel for arbitrary constant $k$.

cs.IT↗
arXiv2026

Spatially-Coupled Network RNA Velocities: A Control-Theoretic Perspective

RNA velocity is an important model that combines cellular spliced and unspliced RNA counts to infer dynamical properties of various regulatory functions. Despite its wide applicability and many variants used in practice, the model has not been adequately designed to directly account for both intracellular gene regulatory network interactions and spatial intercellular communications. Here, we propose a new RNA velocity approach that jointly and directly captures two new network structures: an intracellular gene regulatory network (GRN) and an intercellular interaction network that captures interactions between (neighboring) cells, with relevance to spatial transcriptomics. We theoretically analyze this two-level network system through the lens of control and consensus theory. In particular, we investigate network equilibria, stability, cellular network consensus, and optimal control approaches for targeted drug intervention.

math.DS↗
arXiv2025

The Trojan Knowledge: Bypassing Commercial LLM Guardrails via Harmless Prompt Weaving and Adaptive Tree Search

Large language models (LLMs) remain vulnerable to jailbreak attacks that bypass safety guardrails to elicit harmful outputs. Existing approaches overwhelmingly operate within the prompt-optimization paradigm: whether through traditional algorithmic search or recent agent-based workflows, the resulting prompts typically retain malicious semantic signals that modern guardrails are primed to detect. In contrast, we identify a deeper, largely overlooked vulnerability stemming from the highly interconnected nature of an LLM's internal knowledge. This structure allows harmful objectives to be realized by weaving together sequences of benign sub-queries, each of which individually evades detection. To exploit this loophole, we introduce the Correlated Knowledge Attack Agent (CKA-Agent), a dynamic framework that reframes jailbreaking as an adaptive, tree-structured exploration of the target model's knowledge base. The CKA-Agent issues locally innocuous queries, uses model responses to guide exploration across multiple paths, and ultimately assembles the aggregated information to achieve the original harmful objective. Evaluated across state-of-the-art commercial LLMs (Gemini2.5-Flash/Pro, GPT-oss-120B, Claude-Haiku-4.5), CKA-Agent consistently achieves over 95% success rates even against strong guardrails, underscoring the severity of this vulnerability and the urgent need for defenses against such knowledge-decomposition attacks. Our codes are available at https://github.com/Graph-COM/CKA-Agent.

cs.CR↗
arXiv2025

DMol: A Highly Efficient and Chemical Motif-Preserving Molecule Generation Platform

We introduce a new graph diffusion model for small molecule generation, DMol, which outperforms the state-of-the-art DiGress model in terms of validity by roughly 1.5% across all benchmarking datasets while reducing the number of diffusion steps by at least 10-fold, and the running time to roughly one half. The performance improvements are a result of a careful change in the objective function and a graph noise scheduling approach which, at each diffusion step, allows one to only change a subset of nodes of varying size in the molecule graph. Another relevant property of the method is that it can be easily combined with junction-tree-like graph representations that arise by compressing a collection of relevant ring structures into supernodes. Unlike classical junction-tree techniques that involve VAEs and require complicated reconstruction steps, compressed DMol directly performs graph diffusion on a graph that compresses only a carefully selected set of frequent carbon rings into supernodes, which results in straightforward sample generation. This compressed DMol method offers additional validity improvements over generic DMol of roughly 2%, increases the novelty of the method, and further improves the running time due to reductions in the graph size.

cs.LG↗
arXiv2025

Do LLMs Really Forget? Evaluating Unlearning with Knowledge Correlation and Confidence Awareness

Machine unlearning techniques aim to mitigate unintended memorization in large language models (LLMs). However, existing approaches predominantly focus on the explicit removal of isolated facts, often overlooking latent inferential dependencies and the non-deterministic nature of knowledge within LLMs. Consequently, facts presumed forgotten may persist implicitly through correlated information. To address these challenges, we propose a knowledge unlearning evaluation framework that more accurately captures the implicit structure of real-world knowledge by representing relevant factual contexts as knowledge graphs with associated confidence scores. We further develop an inference-based evaluation protocol leveraging powerful LLMs as judges; these judges reason over the extracted knowledge subgraph to determine unlearning success. Our LLM judges utilize carefully designed prompts and are calibrated against human evaluations to ensure their trustworthiness and stability. Extensive experiments on our newly constructed benchmark demonstrate that our framework provides a more realistic and rigorous assessment of unlearning performance. Moreover, our findings reveal that current evaluation strategies tend to overestimate unlearning effectiveness. Our code is publicly available at https://github.com/Graph-COM/Knowledge_Unlearning.git.

cs.CL↗
arXiv2025

GUARD: Guided Unlearning and Retention via Data Attribution for Large Language Models

Unlearning in large language models is becoming increasingly important due to regulatory compliance, copyright protection, and privacy concerns. However, a key challenge in LLM unlearning is unintended forgetting, where the removal of specific data inadvertently impairs the utility of the model and its retention of valuable, desired information. While prior work has primarily focused on architectural innovations, the influence of data-level factors on unlearning performance remains underexplored. As a result, existing methods often suffer from degraded retention when forgetting high-impact data. To address this problem, we propose GUARD, a novel framework for Guided Unlearning And Retention via Data attribution. At its core, GUARD introduces a lightweight proxy data attribution metric tailored for LLM unlearning, which quantifies the alignment between the Forget and Retain sets while remaining computationally efficient. Building on this, we design a novel unlearning objective that assigns adaptive, nonuniform unlearning weights to samples, inversely proportional to their proxy attribution scores. Through such a reallocation of unlearning power, GUARD mitigates unintended retention loss. We also provide rigorous theoretical guarantees that GUARD significantly improves retention while maintaining forgetting metrics comparable to prior methods. Extensive experiments on the TOFU and MUSE benchmarks across multiple LLM architectures demonstrate that GUARD reduces utility sacrifice on the TOFU Retain Set by up to 194.92 percent in terms of Truth Ratio when forgetting 10 percent of the training data, and improves knowledge retention on the MUSE NEWS Retain Set by 16.20 percent, with comparable or very moderate increases in privacy loss compared to state-of-the-art methods.

cs.LG↗
arXiv2025

DNA Tails for Molecular Flash Memory

DNA-based data storage systems face practical challenges due to the high cost of DNA synthesis. A strategy to address the problem entails encoding data via topological modifications of the DNA sugar-phosphate backbone. The DNA Punchcards system, which introduces nicks (cuts) in the DNA backbone, encodes only one bit per nicking site, limiting density. We propose \emph{DNA Tails,} a storage paradigm that encodes nonbinary symbols at nicking sites by growing enzymatically synthesized single-stranded DNA of varied lengths. The average tail lengths encode multiple information bits and are controlled via a staggered nicking-tail extension process. We demonstrate the feasibility of this encoding approach experimentally and identify common sources of errors, such as calibration errors and stumped tail growth errors. To mitigate calibration errors, we use rank modulation proposed for flash memory. To correct stumped tail growth errors, we introduce a new family of rank modulation codes that can correct ``stuck-at'' errors. Our analytical results include constructions for order-optimal-redundancy permutation codes and accompanying encoding and decoding algorithms.

cs.ET↗
arXiv2025

Generalized Orthogonal de Bruijn and Kautz Sequences

A de Bruijn sequence of order $k$ over a finite alphabet is a cyclic sequence with the property that it contains every possible $k$-sequence as a substring exactly once. Orthogonal de Bruijn sequences are collections of de Bruijn sequences of the same order, $k$, satisfying the joint constraint that every $(k+1)$-sequence appears as a substring in at most one of the sequences in the collection. Both de Bruijn and orthogonal de Bruijn sequences have found numerous applications in synthetic biology, although the latter remain largely unexplored in the coding theory literature. Here we study three relevant practical generalizations of orthogonal de Bruijn sequences where we relax either the constraint that every $(k+1)$-sequence appears exactly once, or that the sequences themselves are de Bruijn rather than balanced de Bruijn sequences. We also provide lower and upper bounds on the number of fixed-weight orthogonal de Bruijn sequences. The paper concludes with parallel results for orthogonal nonbinary Kautz sequences, which satisfy similar constraints as de Bruijn sequences except for only being required to cover all subsequences of length $k$ whose maximum runlength equals to one.

cs.IT↗
arXiv2024

Reducing Data Fragmentation in Data Deduplication Systems via Partial Repetition and Coding

Data deduplication, one of the key features of modern Big Data storage devices, is the process of removing replicas of data chunks stored by different users. Despite the importance of deduplication, several drawbacks of the method, such as storage robustness and file fragmentation, have not been previously analyzed from a theoretical point of view. Storage robustness pertains to ensuring that deduplicated data can be used to reconstruct the original files without service disruptions and data loss. Fragmentation pertains to the problems of placing deduplicated data chunks of different user files in a proximity-preserving linear order, since neighboring chunks of the same file may be stored in sectors far apart on the server. This work proposes a new theoretical model for data fragmentation and introduces novel graph- and coding-theoretic approaches for reducing fragmentation via limited duplication (repetition coding) and coded deduplication (e.g., linear coding). In addition to alleviating issues with fragmentation, limited duplication and coded deduplication can also serve the dual purpose of increasing the robusteness of the system design. The contributions of our work are three-fold. First, we describe a new model for file structures in the form of self-avoiding (simple) paths in specialized graphs. Second, we introduce several new metrics for measuring the fragmentation level in deduplication systems on graph-structured files, including the stretch metric that captures the worst-case "spread" of adjacent data chunks within a file when deduplicated and placed on the server; and, the jump metric that captures the worst-case number of times during the reconstruction process of a file that one has to change the readout location on the server. For the stretch metric, we establish a connection between the level of fragmentation and the bandwidth of the file-graph. In particular, ...

cs.IT↗
arXiv2024

FedGTST: Boosting Global Transferability of Federated Models via Statistics Tuning

The performance of Transfer Learning (TL) heavily relies on effective pretraining, which demands large datasets and substantial computational resources. As a result, executing TL is often challenging for individual model developers. Federated Learning (FL) addresses these issues by facilitating collaborations among clients, expanding the dataset indirectly, distributing computational costs, and preserving privacy. However, key challenges remain unresolved. First, existing FL methods tend to optimize transferability only within local domains, neglecting the global learning domain. Second, most approaches rely on indirect transferability metrics, which do not accurately reflect the final target loss or true degree of transferability. To address these gaps, we propose two enhancements to FL. First, we introduce a client-server exchange protocol that leverages cross-client Jacobian (gradient) norms to boost transferability. Second, we increase the average Jacobian norm across clients at the server, using this as a local regularizer to reduce cross-client Jacobian variance. Our transferable federated algorithm, termed FedGTST (Federated Global Transferability via Statistics Tuning), demonstrates that increasing the average Jacobian and reducing its variance allows for tighter control of the target loss. This leads to an upper bound on the target loss in terms of the source loss and source-target domain discrepancy. Extensive experiments on datasets such as MNIST to MNIST-M and CIFAR10 to SVHN show that FedGTST outperforms relevant baselines, including FedSR. On the second dataset pair, FedGTST improves accuracy by 9.8% over FedSR and 7.6% over FedIIR when LeNet is used as the backbone.

cs.LG↗
arXiv2024

Federated Aggregation of Mallows Rankings: A Comparative Analysis of Borda and Lehmer Coding

Rank aggregation combines multiple ranked lists into a consensus ranking. In fields like biomedical data sharing, rankings may be distributed and require privacy. This motivates the need for federated rank aggregation protocols, which support distributed, private, and communication-efficient learning across multiple clients with local data. We present the first known federated rank aggregation methods using Borda scoring and Lehmer codes, focusing on the sample complexity for federated algorithms on Mallows distributions with a known scaling factor $ϕ$ and an unknown centroid permutation $σ_0$. Federated Borda approach involves local client scoring, nontrivial quantization, and privacy-preserving protocols. We show that for $ϕ\in [0,1)$, and arbitrary $σ_0$ of length $N$, it suffices for each of the $L$ clients to locally aggregate $\max\{C_1(ϕ), C_2(ϕ)\frac{1}{L}\log \frac{N}δ\}$ rankings, where $C_1(ϕ)$ and $C_2(ϕ)$ are constants, quantize the result, and send it to the server who can then recover $σ_0$ with probability $\geq 1-δ$. Communication complexity scales as $NL \log N$. Our results represent the first rigorous analysis of Borda's method in centralized and distributed settings under the Mallows model. Federated Lehmer coding approach creates a local Lehmer code for each client, using a coordinate-majority aggregation approach with specialized quantization methods for efficiency and privacy. We show that for $ϕ+ϕ^2<1+ϕ^N$, and arbitrary $σ_0$ of length $N$, it suffices for each of the $L$ clients to locally aggregate $\max\{C_3(ϕ), C_4(ϕ)\frac{1}{L}\log \frac{N}δ\}$ rankings, where $C_3(ϕ)$ and $C_4(ϕ)$ are constants. Clients send truncated Lehmer coordinate histograms to the server, which can recover $σ_0$ with probability $\geq 1-δ$. Communication complexity is $\sim O(N\log NL\log L)$.

cs.LG↗
arXiv2024

Perturbation-Resilient Trades for Dynamic Service Balancing

A combinatorial trade is a pair of sets of blocks of elements that can be exchanged while preserving relevant subset intersection constraints. The class of balanced and swap-robust minimal trades was proposed in [1] for exchanging blocks of data chunks stored on distributed storage systems in an access- and load-balanced manner. More precisely, data chunks in the trades of interest are labeled by popularity ranks and the blocks are required to have both balanced overall popularity and stability properties with respect to swaps in chunk popularities. The original construction of such trades relied on computer search and paired balanced sets obtained through iterative combining of smaller sets that have provable stability guarantees. To reduce the substantial gap between the results of prior approaches and the known theoretical lower bound, we present new analytical upper and lower bounds on the minimal disbalance of blocks introduced by limited-magnitude popularity ranking swaps. Our constructive and near-optimal approach relies on pairs of graphs whose vertices are two balanced sets with edges/arcs that capture the balance and potential balance changes induced by limited-magnitude popularity swaps. In particular, we show that if we start with carefully selected balanced trades and limit the magnitude of rank swaps to one, the new upper and lower bound on the maximum block disbalance caused by a swap only differ by a factor of $1.07$. We also extend these results for larger popularity swap magnitudes.

cs.DS↗
arXiv2024

Graph Transductive Defense: a Two-Stage Defense for Graph Membership Inference Attacks

Graph neural networks (GNNs) have become instrumental in diverse real-world applications, offering powerful graph learning capabilities for tasks such as social networks and medical data analysis. Despite their successes, GNNs are vulnerable to adversarial attacks, including membership inference attacks (MIA), which threaten privacy by identifying whether a record was part of the model's training data. While existing research has explored MIA in GNNs under graph inductive learning settings, the more common and challenging graph transductive learning setting remains understudied in this context. This paper addresses this gap and proposes an effective two-stage defense, Graph Transductive Defense (GTD), tailored to graph transductive learning characteristics. The gist of our approach is a combination of a train-test alternate training schedule and flattening strategy, which successfully reduces the difference between the training and testing loss distributions. Extensive empirical results demonstrate the superior performance of our method (a decrease in attack AUROC by $9.42\%$ and an increase in utility performance by $18.08\%$ on average compared to LBP), highlighting its potential for seamless integration into various classification models with minimal overhead.

cs.LG↗
arXiv2024

Multiple sequences Prophet Inequality Under Observation Constraints

In our problem, we are given access to a number of sequences of nonnegative i.i.d. random variables, whose realizations are observed sequentially. All sequences are of the same finite length. The goal is to pick one element from each sequence in order to maximize a reward equal to the expected value of the sum of the selections from all sequences. The decision on which element to pick is irrevocable, i.e., rejected observations cannot be revisited. Furthermore, the procedure terminates upon having a single selection from each sequence. Our observation constraint is that we cannot observe the current realization of all sequences at each time instant. Instead, we can observe only a smaller, yet arbitrary, subset of them. Thus, together with a stopping rule that determines whether we choose or reject the sample, the solution requires a sampling rule that determines which sequence to observe at each instant. The problem can be solved via dynamic programming, but with an exponential complexity in the length of the sequences. In order to make the solution computationally tractable, we introduce a decoupling approach and determine each stopping time using either a single-sequence dynamic programming, or a Prophet Inequality inspired threshold method, with polynomial complexity in the length of the sequences. We prove that the decoupling approach guarantees at least 0.745 of the optimal expected reward of the joint problem. In addition, we describe how to efficiently compute the optimal number of samples for each sequence, and its' dependence on the variances.

math.ST↗
arXiv2024

Online Distribution Learning with Local Private Constraints

We study the problem of online conditional distribution estimation with \emph{unbounded} label sets under local differential privacy. Let $\mathcal{F}$ be a distribution-valued function class with unbounded label set. We aim at estimating an \emph{unknown} function $f\in \mathcal{F}$ in an online fashion so that at time $t$ when the context $\boldsymbol{x}_t$ is provided we can generate an estimate of $f(\boldsymbol{x}_t)$ under KL-divergence knowing only a privatized version of the true labels sampling from $f(\boldsymbol{x}_t)$. The ultimate objective is to minimize the cumulative KL-risk of a finite horizon $T$. We show that under $(ε,0)$-local differential privacy of the privatized labels, the KL-risk grows as $\tildeΘ(\frac{1}ε\sqrt{KT})$ upto poly-logarithmic factors where $K=|\mathcal{F}|$. This is in stark contrast to the $\tildeΘ(\sqrt{T\log K})$ bound demonstrated by Wu et al. (2023a) for bounded label sets. As a byproduct, our results recover a nearly tight upper bound for the hypothesis selection problem of gopi et al. (2020) established only for the batch setting.

cs.LG↗
arXiv2024

Federated Classification in Hyperbolic Spaces via Secure Aggregation of Convex Hulls

Hierarchical and tree-like data sets arise in many applications, including language processing, graph data mining, phylogeny and genomics. It is known that tree-like data cannot be embedded into Euclidean spaces of finite dimension with small distortion. This problem can be mitigated through the use of hyperbolic spaces. When such data also has to be processed in a distributed and privatized setting, it becomes necessary to work with new federated learning methods tailored to hyperbolic spaces. As an initial step towards the development of the field of federated learning in hyperbolic spaces, we propose the first known approach to federated classification in hyperbolic spaces. Our contributions are as follows. First, we develop distributed versions of convex SVM classifiers for Poincaré discs. In this setting, the information conveyed from clients to the global classifier are convex hulls of clusters present in individual client data. Second, to avoid label switching issues, we introduce a number-theoretic approach for label recovery based on the so-called integer $B_h$ sequences. Third, we compute the complexity of the convex hulls in hyperbolic spaces to assess the extent of data leakage; at the same time, in order to limit communication cost for the hulls, we propose a new quantization method for the Poincaré disc coupled with Reed-Solomon-like encoding. Fourth, at the server level, we introduce a new approach for aggregating convex hulls of the clients based on balanced graph partitioning. We test our method on a collection of diverse data sets, including hierarchical single-cell RNA-seq data from different patients distributed across different repositories that have stringent privacy constraints. The classification accuracy of our method is up to $\sim 11\%$ better than its Euclidean counterpart, demonstrating the importance of privacy-preserving learning in hyperbolic spaces.

cs.LG↗