arXiv subjects

Duy Le

Publications and source records attributed to Duy Le.

17 recordsLinked to original sources

arXiv2025

Strategic Server Deployment under Uncertainty in Mobile Edge Computing

Server deployment is a fundamental task in mobile edge computing: where to place the edge servers and what user cells to assign to them. To make this decision is context-specific, but common goals are 1) computing efficiency: maximize the amount of workload processed by the edge, and 2) communication efficiency: minimize the communication cost between the cells and their assigned servers. We focus on practical scenarios where the user workload in each cell is unknown and time-varying, and so are the effective capacities of the servers. Our research problem is to choose a subset of candidate servers and assign them to the user cells such that the above goals are sustainably achieved under the above uncertainties. We formulate this problem as a stochastic bilevel optimization, which is strongly NP-hard and unseen in the literature. By approximating the objective function with submodular functions, we can utilize state-of-the-art greedy algorithms for submodular maximization to effectively solve our problem. We evaluate the proposed algorithm using real-world data, showing its superiority to alternative methods; the improvement can be as high as 55%

cs.DC↗

arXiv2025

Machine learned potential for defected single layer hexagonal boron nitride

Development of machine learned interatomic potentials (MLIP) is critical for performing reliable simulations of materials at length and time scales that are comparable to those in the laboratory. We present here a MLIP suitable for simulations of the temperature dependent structure and dynamics of single layer hexagonal boron nitride (h-BN) with defects and grain boundaries, developed using a strictly local equivariant deep neural network as formulated in the Allegro code. The training dataset consisted of about 30,000 images of h-BN with and without point defects generated with ab-initio molecular dynamics simulations, based on density functional theory (DFT), at 500, 1000, and 1500K. The developed MLIP predicts potential energies and forces with a mean absolute error (MAE) of 4 meV/atom and 60 meV/Angstrom , respectively. It also reproduces phonon dispersion curves and density of vibrational states of pristine bulk h-BN that are comparable with that obtained from density functional theory-based calculations. Molecular dynamics simulations of the motion of the 4|8 grain boundary unit in h-BN shows the first step to have an activation barrier ~2.2 eV, indicating immobility of the grain boundary. Moving the grain boundary units past the first shows much lower activation barriers of ~0.42eV, suggesting a facile motion of the grain boundary once the first movement is stimulated. These simulations yield a scaled mobility of 1.739*10^(-11) m^3/Js for a temperature of 1500K which, given the inherent differences in the set-ups, is not too far from the experimental value of 1.36*10^(-9) m^3/Js. The ability to predict grain boundary mobility within reasonable agreement with experiment demonstrates the robustness of the MLIP and its suitability for reliable simulations of defect structures and dynamics in single layer h-BN.

cond-mat.mtrl-sci↗

arXiv2025

RobustFSM: Submodular Maximization in Federated Setting with Malicious Clients

Submodular maximization is an optimization problem benefiting many machine learning applications, where we seek a small subset best representing an extremely large dataset. We focus on the federated setting where the data are locally owned by decentralized clients who have their own definitions for the quality of representability. This setting requires repetitive aggregation of local information computed by the clients. While the main motivation is to respect the privacy and autonomy of the clients, the federated setting is vulnerable to client misbehaviors: malicious clients might share fake information. An analogy is backdoor attack in conventional federated learning, but our challenge differs freshly due to the unique characteristics of submodular maximization. We propose RobustFSM, a federated submodular maximization solution that is robust to various practical client attacks. Its performance is substantiated with an empirical evaluation study using real-world datasets. Numerical results show that the solution quality of RobustFSM substantially exceeds that of the conventional federated algorithm when attacks are severe. The degree of this improvement depends on the dataset and attack scenarios, which can be as high as 200%

cs.LG↗

arXiv2025

Adaptive Originality Filtering: Rejection Based Prompting and RiddleScore for Culturally Grounded Multilingual Riddle Generation

Language models are increasingly tested on multilingual creativity, demanding culturally grounded, abstract generations. Standard prompting methods often produce repetitive or shallow outputs. We introduce Adaptive Originality Filtering (AOF), a prompting strategy that enforces novelty and cultural fidelity via semantic rejection. To assess quality, we propose RiddleScore, a metric combining novelty, diversity, fluency, and answer alignment. AOF improves Distinct-2 (0.915 in Japanese), reduces Self-BLEU (0.177), and raises RiddleScore (up to +57.1% in Arabic). Human evaluations confirm fluency, creativity, and cultural fit gains. However, improvements vary: Arabic shows greater RiddleScore gains than Distinct-2; Japanese sees similar changes. Though focused on riddles, our method may apply to broader creative tasks. Overall, semantic filtering with composite evaluation offers a lightweight path to culturally rich generation without fine-tuning.

cs.CL↗

arXiv2025

PFCS: Prime Factorization Cache System for Deterministic Data Relationship Discovery

Cache systems fundamentally limit modern computing performance due to their inability to precisely capture data relationships. While achieving 85-92% hit rates, traditional systems rely on statistical heuristics that cannot guarantee relationship discovery, leading to suboptimal prefetching and resource waste. We present PFCS (Prime Factorization Cache System), which leverages the mathematical uniqueness of prime factorization to achieve deterministic relationship discovery with zero false positives. PFCS assigns unique primes to data elements and represents relationships as composite numbers, enabling the recovery of perfect relationships through factorization. A comprehensive evaluation across database, ML, and HPC workloads demonstrates an average performance improvement of x 6.2, 98.9% hit rates, and a 38% power reduction compared to state-of-the-art systems. The mathematical foundation provides formal guarantees impossible with approximation-based approaches, establishing a new paradigm for cache system design

cs.DB↗

arXiv2025

Unveiling Coverage Dependent Interactions of N-Methylaniline with the Pt(111) Surface

This study aims to elucidate the adsorption and surface chemistry of N-methylaniline (NMA) on Pt(111), using it as a model molecule to probe the activation mechanisms of aromatic amines on catalytic surfaces. Through a combination of density functional theory (DFT) calculations and experimental techniques such as temperature programmed X-ray photoelectron spectroscopy (TP-XPS), temperature programmed desorption (TPD), and Fourier transform infrared reflection absorption spectroscopy(FT-IRRAS),we explored the coverage-dependent behaviour of NMA on Pt(111) to identify key steps in the activation process. The population of certain reaction paths is driven by a coverage dependent balance between molecule surface charge transfer and intermolecular interactions, dictating the selective activation of specific bonds. Our findings reveal how coverage influences the orientation and bonding of NMA on the Pt(111)surface. At lower coverages, the molecule binds to the surface through the phenyl ring and activation, facilitating C-N bond cleavage to the ring under HCN formation. In comparison, at higher coverages, the molecule binds only through the nitrogen atom and desorbs intact. These insights into variable bond activation lay the ground work for understanding the fundamental processes involved in potential heterogeneously catalyzed reactions of aromatic amines, contributing to the development of new catalytic strategies.

cond-mat.mtrl-sci↗

arXiv2024

Sentiment Reasoning for Healthcare

Transparency in AI healthcare decision-making is crucial. By incorporating rationales to explain reason for each predicted label, users could understand Large Language Models (LLMs)'s reasoning to make better decision. In this work, we introduce a new task - Sentiment Reasoning - for both speech and text modalities, and our proposed multimodal multitask framework and the world's largest multimodal sentiment analysis dataset. Sentiment Reasoning is an auxiliary task in sentiment analysis where the model predicts both the sentiment label and generates the rationale behind it based on the input transcript. Our study conducted on both human transcripts and Automatic Speech Recognition (ASR) transcripts shows that Sentiment Reasoning helps improve model transparency by providing rationale for model prediction with quality semantically comparable to humans while also improving model's classification performance (+2% increase in both accuracy and macro-F1) via rationale-augmented fine-tuning. Also, no significant difference in the semantic quality of generated rationales between human and ASR transcripts. All code, data (five languages - Vietnamese, English, Chinese, German, and French) and models are published online: https://github.com/leduckhai/Sentiment-Reasoning

cs.CL↗

arXiv2024

KV Cache Compression, But What Must We Give in Return? A Comprehensive Benchmark of Long Context Capable Approaches

Long context capability is a crucial competency for large language models (LLMs) as it mitigates the human struggle to digest long-form texts. This capability enables complex task-solving scenarios such as book summarization, code assistance, and many more tasks that are traditionally manpower-intensive. However, transformer-based LLMs face significant challenges with long context input due to the growing size of the KV cache and the intrinsic complexity of attending to extended inputs; where multiple schools of efficiency-driven approaches - such as KV cache quantization, token dropping, prompt compression, linear-time sequence models, and hybrid architectures - have been proposed to produce efficient yet long context-capable models. Despite these advancements, no existing work has comprehensively benchmarked these methods in a reasonably aligned environment. In this work, we fill this gap by providing a taxonomy of current methods and evaluating 10+ state-of-the-art approaches across seven categories of long context tasks. Our work reveals numerous previously unknown phenomena and offers insights - as well as a friendly workbench - for the future development of long context-capable LLMs. The source code is available at https://github.com/henryzhongsc/longctx_bench.

cs.CL↗

arXiv2023

ZEETAD: Adapting Pretrained Vision-Language Model for Zero-Shot End-to-End Temporal Action Detection

Temporal action detection (TAD) involves the localization and classification of action instances within untrimmed videos. While standard TAD follows fully supervised learning with closed-set setting on large training data, recent zero-shot TAD methods showcase the promising open-set setting by leveraging large-scale contrastive visual-language (ViL) pretrained models. However, existing zero-shot TAD methods have limitations on how to properly construct the strong relationship between two interdependent tasks of localization and classification and adapt ViL model to video understanding. In this work, we present ZEETAD, featuring two modules: dual-localization and zero-shot proposal classification. The former is a Transformer-based module that detects action events while selectively collecting crucial semantic embeddings for later recognition. The latter one, CLIP-based module, generates semantic embeddings from text and frame inputs for each temporal unit. Additionally, we enhance discriminative capability on unseen classes by minimally updating the frozen CLIP encoder with lightweight adapters. Extensive experiments on THUMOS14 and ActivityNet-1.3 datasets demonstrate our approach's superior performance in zero-shot TAD and effective knowledge transfer from ViL models to unseen action categories.

cs.CV↗

arXiv2023

Thickness dependence of superconductivity in FeSe films

The films of FeSe on substrates have attracted attention because of their unusually high-temperature (Tc) superconducting properties whose origins continue to be debated. To disentangle the competing effects of the substrate and interlayer and intralayer processes, we present here results of density functional theory (DFT)-based analysis of the electronic structure of unsupported FeSe films consisting of 1 to 5 layers (1L-5L). Furthermore, by solving the Bardeen-Schrieffer-Cooper (BCS) equation with spin-wave exchange attraction derived from the Hubbard model, we find the superconducting critical temperature Tc for 1L-5L and bulk FeSe systems in reasonable agreement with experimental data. Our results point to the importance of correlation effects in superconducting properties of single- and multi-layer FeSe films, independently of the role of substrate.

cond-mat.supr-con↗

arXiv2023

Dark exciton energy splitting in monolayer WSe2: insights from time-dependent density-functional theory

We present here a formalism based on time-dependent density-functional theory (TDDFT) to describe characteristics of both intra- and inter-valley excitons in semiconductors, the latter of which had remained a challenge. Through the usage of an appropriate exchange-correlation kernel (nanoquanta), we trace the energy difference between the intra- and inter-valley dark excitons in monolayer (1L) WSe2 to the domination of the exchange part in the exchange-correlation energies of these states. Furthermore, our calculated transition contribution maps establish the momentum resolved weights of the electron-hole excitations in both bright and dark excitons thereby providing a comprehensive understanding of excitonic properties of 1L WSe2. We find that the states consist of hybridized excitations around the corresponding valleys which leads to brightening of the dark excitons, i.e., significantly decreasing their lifetime which is reflected in the PL spectrum. Using many-body perturbation theory, we calculate the phonon contribution to the energy bandgap and the linewidths of the excited electrons, holes and (bright) exciton to find that as the temperature increases the bandgap significantly decreases, while the linewidths increase. Our work paves for describing the ultrafast charge dynamics of transition metal dichalcogenide within an ab initio framework.

cond-mat.mtrl-sci↗

arXiv2022

Thickness-dependent spin bistable transitions in single-crystalline molecular 2D material

The advent of two-dimensional (2D) crystals has led to numerous scientific breakthroughs. Conventional 2D systems have in-plane covalent bonds and a weak out-of-plane van-der-Waals bond. Here we report a new type of 2D material composed of discrete magnetic molecules, where anisotropic van-der-Waals interactions bond the molecules into a 2D packing. Through mechanical exfoliation, we can obtain single-crystalline molecular monolayers, which can be readily integrated into other 2D systems. Optical spectroscopy suggests the few-layered molecules preserve the temperature-induced spin-crossover switching observed in the bulk form but show a drastic increase in thermal hysteresis unique to these thin 2D molecule assemblies. The trapping of spin bistability with decreasing layer number can arise from domain wall dynamics in reduced dimensions. Our results establish molecular solids with strong anisotropy of intermolecular interactions as precursors to a novel class of 2D materials, affording new possibilities to control functionalities through substrate and interlayer interactions.

cond-mat.mes-hall↗

arXiv2021

Fermi surfaces of the topological semimetal CaSn$_{3}$ probed through de Haas van Alphen oscillations

In the search of topological superconductors, nailing down the Fermiology of the normal state is as crucial a prerequisite as unraveling the superconducting pairing symmetry. In particular, the number of time-reversal-invariant momenta in the Brillouin zone enclosed by Fermi surfaces is closely linked to the topological class of time-reversal-invariant systems, and can experimentally be investigated. We report here a detailed study of de Haas van Alphen quantum oscillations in single crystals of the topological semimetal CaSn$_{3}$ with torque magnetometry in high magnetic fields up to 35 T. In conjunction with density functional theory based calculations, the observed quantum oscillations frequencies indicate that the Fermi surfaces of CaSn$_{3}$ enclose an odd number of time-reversal-invariant momenta, satisfying one of the proposed criteria to realize topological superconductivity. Nonzero Berry phases extracted from the magnetic oscillations also support the nontrivial topological nature of CaSn$_{3}$.

cond-mat.supr-con↗

arXiv2019

Sub-monolayer structures of Ag overlayers on Ge(111): experimental observations and first-principles study

We present a joint experimental and theoretical determination of structures of Ag adatoms on the Ge(111) surface using low energy electron diffraction, low energy electron microscopy, scanning tunneling microscopy, and density functional theory-based calculations, as functions of coverages and temperature. Experimentally for clean Ge(111), c(2X8) and (2X1) phases occur, while Ag overlayers cause (4X4), (V3XV3)R30 and (3X1) surface structural phases. The dependence of the growth behavior of these different phases was examined as a function of temperature, Ag deposition rate and coverage, substrate step density, and history of temperature cycling. First-principles calculations of the electronic and geometric structures and vibrational dynamics show the Ge(111)-c(2X8) configuration with Ge adatoms adsorbed on three-fold hollow (T4) sites to be the energetically most favored phase of the Ge(111) surface, among unreconstructed Ge(111), reconstructed Ge(111)-2X1, and Ge(111)-c(2X8) structures. The Ge(111)-Ag(3X1) overlayer of the system has Ge atoms forming a honeycomb chain on a missing top layer reconstructed surface, with metal at 1/3 ML coverage in channel. The Ge (111)-Ag(V3XV3)R30 overlayer contains one monolayer Ag forming inequivalent Ag triangles in a surface unit cell on the missing top layer reconstructed Ge(111) surface. The Ge(111)-Ag(4X4) overlayer formed at low Ag coverage contains two triangular subunits at different heights: one with six Ag adatoms and the other with three Ge adatoms on the intact double layer Ge(111) surface. The temperature and coverage dependent surface phase diagram, obtained by minimizing the surface free energy, captures the main features of the experimental phase diagram.

cond-mat.mtrl-sci↗

arXiv2016

Disorder effect on the anisotropic resistivity of phosphorene determined by a tight-binding model

In this work we develop a compact multi-orbital tight-binding model for phosphorene that accurately describes states near the main band gap. The model parameters are adjusted using as reference the band structure obtained by a density-functional theory calculation with the hybrid HSE06 functional. We use the optimized tight-binding model to study the effects of disorder on the anisotropic transport properties of phosphorene. In particular, we evaluate how the longitudinal resistivity depends on the lattice orientation for two typical disorder models: dilute scatterers with high potential fluctuation amplitudes, mimicking screened charges in the substrate, and dense scatterers with lower amplitudes, simulating weakly bounded adsorbates. We show that the intrinsic anisotropy associated to the band structure of this material, although sensitive to the type and intensity of the disorder, is robust.

cond-mat.mes-hall↗

arXiv2015

Strength of the dominant scatterer in graphene on silicon oxide

A large variability of carrier mobility of graphene-based field effect transistors hampers graphene science and technology. We determine the scattering strength of the dominant scatterer responsible for the variability of graphene-based transistors on silicon oxide. The strength of the scatterer is found to be more consistent with charged impurities than with resonant impurities.

cond-mat.mes-hall↗

arXiv2015

Electron-phonon interaction, excitations and ultrafast photoemission from doped monolayer MoS2

We analyze the effect of electron-phonon coupling on photoemission properties and ultrafast response of doped monolayer MoS2. The analysis is based on combined DFT and many-body (Eliashberg theory) approaches. In particular, we have calculated the electronic and phonon spectra, the electron-phonon coupling and the electronic spectral function of the system at different values of doping. We have also analyzed the emissive properties and the response of the system to femtosecond (fs) laser pulses. It is shown that position of the emission peak of undoped system is in agreement with the experimental data if one takes into account the excitonic effects. The results for the self-energy and spectral functions of the doped systems suggest that one can expect ultrafast processes to be important in the system response , which makes the system attractive from the point of view of modern technological applications. Similar to graphene, the doped system demonstrates ultrafast (fs) relaxation of the electronic subsystem when excited by fs pulses, and a high ultrafast phonon relaxation-induced spectral fluence of visible light emission. Together with high carrier mobility, these features of monolayer MoS2 might be used in modern optoelectronic technologies.

cond-mat.mtrl-sci↗