Search

The impact of voice disorders is becoming more widely acknowledged as a public health issue. Several machine learning-based classifiers with the potential to identify disorders have been used in recent studies to differentiate between normal and pathological voices and sounds. In this paper, we focus on analyzing the vulnerabilities of these systems by exploring the possibility of attacks that can reverse classification and compromise their reliability. Given the critical nature of personal health information, understanding which types of attacks are effective is a necessary first step toward improving the security of such systems. Starting from the original audios, we implement various attack methods, including adversarial, evasion, and pitching techniques, and evaluate how state-of-the-art disorder detection models respond to them. Our findings identify the most effective attack strategies, underscoring the need to address these vulnerabilities in machine-learning systems used in the healthcare domain., Comment: 7 pages, 17 figures, accepted for 16th IEEE INTERNATIONAL WORKSHOP ON INFORMATION FORENSICS AND SECURITY (WIFS) 2024

Hate speech is one of the main threats posed by the widespread use of social networks, despite efforts to limit it. Although attention has been devoted to this issue, the lack of datasets and case studies centered around scarcely represented phenomena, such as ableism or ageism, can lead to hate speech detection systems that do not perform well on underrepresented identity groups. Given the unpreceded capabilities of LLMs in producing high-quality data, we investigate the possibility of augmenting existing data with generative language models, reducing target imbalance. We experiment with augmenting 1,000 posts from the Measuring Hate Speech corpus, an English dataset annotated with target identity information, adding around 30,000 synthetic examples using both simple data augmentation methods and different types of generative models, comparing autoregressive and sequence-to-sequence approaches. We find traditional DA methods to often be preferable to generative models, but the combination of the two tends to lead to the best results. Indeed, for some hate categories such as origin, religion, and disability, hate speech classification using augmented data for training improves by more than 10% F1 over the no augmentation baseline. This work contributes to the development of systems for hate speech detection that are not only better performing but also fairer and more inclusive towards targets that have been neglected so far.

The anisotropic potential energy surface of the (H$_2$)$_2$ dimer represents a challenging problem for many-body methods. Here, we determine the potential energy curves of five different dimer configurations (T, Z, X, H, L) using the lattice regularized diffusion Monte Carlo (LRDMC) method and a number of approximate functionals within density functional theory (DFT), including advanced orbital-dependent functionals based on the random phase approximation (RPA). We assess their performance in describing the potential wells, bond distances and relative energies. The repulsive potential wall is studied by looking at the relative stability of the different dimer configurations as a function of an applied force acting along the intermolecular axis. It is shown that most functionals within DFT break down at finite compression, even those that give an accurate description around the potential well minima. Only by including exchange within RPA a qualitatively correct description along the entire potential energy curve is obtained. Finally, we discuss these results in the context of solid molecular hydrogen at finite pressures., Comment: 25 pages, 15 figures

3D face reconstruction (3DFR) algorithms are based on specific assumptions tailored to distinct application scenarios. These assumptions limit their use when acquisition conditions, such as the subject's distance from the camera or the camera's characteristics, are different than expected, as typically happens in video surveillance. Additionally, 3DFR algorithms follow various strategies to address the reconstruction of a 3D shape from 2D data, such as statistical model fitting, photometric stereo, or deep learning. In the present study, we explore the application of three 3DFR algorithms representative of the SOTA, employing each one as the template set generator for a face verification system. The scores provided by each system are combined by score-level fusion. We show that the complementarity induced by different 3DFR algorithms improves performance when tests are conducted at never-seen-before distances from the camera and camera characteristics (cross-distance and cross-camera settings), thus encouraging further investigations on multiple 3DFR-based approaches., Comment: Accepted at T-CAP - Towards a Complete Analysis of People: Fine-grained Understanding for Real-World Applications, workshop in conjunction with the 18th European Conference on Computer Vision ECCV 2024

In the rapidly evolving field of online fashion shopping, the need for more personalized and interactive image retrieval systems has become paramount. Existing methods often struggle with precisely manipulating specific garment attributes without inadvertently affecting others. To address this challenge, we propose GAMMA (Garment Attribute Manipulation with Multi-level Attention), a novel framework that integrates attribute-disentangled representations with a multi-stage attention-based architecture. GAMMA enables targeted manipulation of fashion image attributes, allowing users to refine their searches with high accuracy. By leveraging a dual-encoder Transformer and memory block, our model achieves state-of-the-art performance on popular datasets like Shopping100k and DeepFashion., Comment: Accepted for publication at the ECCV 2024 workshop FashionAI

Exploring and understanding language data is a fundamental stage in all areas dealing with human language. It allows NLP practitioners to uncover quality concerns and harmful biases in data before training, and helps linguists and social scientists to gain insight into language use and human behavior. Yet, there is currently a lack of a unified, customizable tool to seamlessly inspect and visualize language variation and bias across multiple variables, language units, and diverse metrics that go beyond descriptive statistics. In this paper, we introduce Variationist, a highly-modular, extensible, and task-agnostic tool that fills this gap. Variationist handles at once a potentially unlimited combination of variable types and semantics across diversity and association metrics with regards to the language unit of choice, and orchestrates the creation of up to five-dimensional interactive charts for over 30 variable type-semantics combinations. Through our case studies on computational dialectology, human label variation, and text generation, we show how Variationist enables researchers from different disciplines to effortlessly answer specific research questions or unveil undesired associations in language data. A Python library, code, documentation, and tutorials are made publicly available to the research community., Comment: ACL 2024 (System Demonstrations)

We investigate the structural and thermodynamic properties of high-pressure ice by incorporating quantum anharmonicity at a non-perturbative level. Quantum fluctuations reduce the critical pressure of the phase transition between phase VIII (with asymmetric H-bonds) and phase X (with symmetric H-bonds) by 65 GPa from its classical value of 116 GPa at 0K. Moreover, quantum effects make it temperature-independent over a wide temperature range (0K-300K), in agreement with experimental estimates obtained through vibrational spectroscopy and in striking contrast to the strong temperature dependence found in the classical approximation. The equation of state shows fingerprints of the transition in accordance with experimental evidence. Additionally, we demonstrate that, within our approach, proton disorder in phase VII has a negligible impact on the occurrence of phase X. Finally, we reproduce with high accuracy the 10 GPa isotope shift due to the hydrogen-to-deuterium substitution.

Ab initio quantum Monte Carlo (QMC) is a state-of-the-art numerical approach for evaluating accurate expectation values of many-body wavefunctions. However, one of the major drawbacks that still hinders widespread QMC applications is the lack of an affordable scheme to compute unbiased atomic forces. In this study, we propose a very efficient method to obtain unbiased atomic forces and pressures in the Variational Monte Carlo (VMC) framework with the Jastrow-correlated Slater determinant ansatz, exploiting the gauge-invariant and locality properties of its geminal representation. We demonstrate the effectiveness of our method for H$_2$ and Cl$_2$ molecules and for the cubic boron nitride crystal. Our framework has a better algorithmic scaling with the system size than the traditional finite-difference method, and, in practical applications, is as efficient as single-point VMC calculations. Thus, it paves the way to study dynamical properties of materials, such as phonons, and is beneficial for pursuing more reliable machine-learning interatomic potentials based on unbiased VMC forces., Comment: 7 pages, 5 figures

In the quest of new exotic phases of matter due to the interplay of various interactions, iridates hosting a spin-orbit entangled $j_{\mathrm{eff}}=1/2$ ground state have been in the spotlight in recent years. Also in view of parallels with the low-energy physics of high-temperature superconducting cuprates, the validity of a single- or few-band picture in terms of the $j_{\mathrm{eff}}$ states is key. However, in particular for its structurally simple member Ba$_2$IrO$_4$, such a systematic construction and subsequent analysis of minimal low-energy models are still missing. Here we show by means of a combination of different ab initio techniques with dynamical mean-field theory that a three-band model in terms of Ir-$j_{\mathrm{eff}}$ states fully retains the low-energy physics of the system as compared to a full Ir-$5d$ model. Providing a detailed study of the three-band model in terms of spin-orbit coupling, Hund's coupling and Coulomb interactions, we map out a rich phase diagram and identify a region of effective one-band metal-insulator transition relevant to Ba$_2$IrO$_4$. Compared to available angle-resolved photoemission spectra, we find good agreement of salient aspects of the calculated spectral function and identify features which require the inclusion of non-local fluctuations. In a broader context, we envisage the three- and five-band models developed in this study to be relevant for the study of doped Ba$_2$IrO$_4$ and to clarify further the similarities and differences with cuprates., Comment: 14+1 pages, 13+2 figures, 3+1 tables; comments are welcome

TurboGenius is an open-source Python package designed to fully control ab initio quantum Monte Carlo (QMC) jobs using a Python script, which allows one to perform high-throughput calculations combined with TurboRVB [K. Nakano et al. J. Phys. Chem. 152, 204121 (2020)]. This paper provides an overview of the TurboGenius package and showcases several results obtained in a high-throughput mode. For the purpose of performing high-throughput calculations with TurboGenius, we implemented another open-source Python package, TurboWorkflows, that enables one to construct simple workflows using TurboGenius. We demonstrate its effectiveness by performing (1) validations of density functional theory (DFT) and QMC drivers as implemented in the TurboRVB package and (2) benchmarks of Diffusion Monte Carlo (DMC) calculations for several data sets. For (1), we checked inter-package consistencies between TurboRVB and other established quantum chemistry packages. By doing so, we confirmed that DFT energies obtained by PySCF are consistent with those obtained by TurboRVB within the local density approximation (LDA), and that Hartree-Fock (HF) energies obtained by PySCF and Quantum Package are consistent with variational Monte Carlo energies obtained by TurboRVB with the HF wavefunctions. These validation tests constitute a further reliability check of the TurboRVB package. For (2), we benchmarked atomization energies of the Gaussian-2 set, binding energies of the S22, A24, and SCAI sets, and equilibrium lattice parameters of 12 cubic crystals using DMC calculations. We found that, for all compounds analyzed here, the DMC calculations with the LDA nodal surface give satisfactory results, i.e., consistent either with high-level computational or with experimental reference values., Comment: 44 pages

The International Fingerprint Liveness Detection Competition (LivDet) is a biennial event that invites academic and industry participants to prove their advancements in Fingerprint Presentation Attack Detection (PAD). This edition, LivDet2023, proposed two challenges, Liveness Detection in Action and Fingerprint Representation, to evaluate the efficacy of PAD embedded in verification systems and the effectiveness and compactness of feature sets. A third, hidden challenge is the inclusion of two subsets in the training set whose sensor information is unknown, testing participants ability to generalize their models. Only bona fide fingerprint samples were provided to participants, and the competition reports and assesses the performance of their algorithms suffering from this limitation in data availability., Comment: 9 pages, 10 tables, IEEE International Joint Conference on Biometrics (IJCB 2023)

We study the structural phase transition, originally associated with the highest superconducting critical temperature $T_c$ measured in high-pressure sulfur hydride. A quantitative description of its pressure dependence has been elusive for any \emph{ab initio} theory attempted so far, raising questions on the actual mechanism leading to the maximum of $T_c$. Here, we estimate the critical pressure of the hydrogen bond symmetrization in the Im$\bar{3}$m structure, by combining density functional theory and quantum Monte Carlo simulations for electrons with path integral molecular dynamics for quantum nuclei. We find that the $T_c$ maximum corresponds to pressures where local dipole moments dynamically form on the hydrogen sites, as precursors of the ferroelectric Im$\bar{3}$m-R3m transition, happening at lower pressures. For comparison, we also apply the self-consistent harmonic approximation, whose ferroelectric critical pressure lies in between the ferroelectric transition estimated by path integral molecular dynamics and the local dipole formation. Nuclear quantum effects play a major role in a significant reduction ($\approx$ 50 GPa) of the classical ferroelectric transition pressure at 200K and in a large isotope shift ($\approx$ 25 GPa) upon hydrogen-to-deuterium substitution of the local dipole formation pressure, in agreement with the corresponding change in the $T_c$ maximum location.

We present a robust reciprocal-space implementation of the temperature-dependent effective potential method. Our implementation can scale easily to large cell and long sampling time. It is interoperable with standard ab-initio molecular dynamics and with Langevin dynamics. We prove that both sampling methods can be efficient and accurate if a thermostat is used to control temperature and dynamics parameters are used to optimize the sampling efficiency. By way of example, we apply it to study anharmonic phonon renormalization in weakly and strongly anharmonic materials, reproducing the temperature effect on phonon frequencies, crossing of phase transition, and stabilization of high-temperature phases.

The interplay of quantum and classical fluctuations in the vicinity of a quantum critical point (QCP) gives rise to various regimes or phases with distinct quantum character. In this work, we show that the R\'enyi entropy is a precious tool to characterize the phase diagram of critical systems not only around the QCP but also away from it, thanks to its capability to detect the emergence of local order at finite temperature. For an efficient evaluation of the R\'enyi entropy, we introduce a new algorithm based on a path integral Langevin dynamics combined with a previously proposed thermodynamic integration method built on regularized paths. We apply this framework to study the critical behavior of a linear chain of anharmonic oscillators, a particular realization of the $\phi^4$ model. We fully resolved its phase diagram, as a function of both temperature and interaction strength. At finite temperature, we find a sequence of three regimes - para, disordered and quasi long-range ordered -, met as the interaction is increased. The R\'enyi entropy divergence coincides with the crossover between the para and disordered regime, which shows no temperature dependence. The occurrence of quasi long-range order, on the other hand, is temperature dependent. The two crossover lines merge in proximity of the QCP, at zero temperature, where the R\'enyi entropy is sharply peaked. Via its subsystem-size scaling, we confirm that the transition belongs to the two-dimensional Ising universality class. This phenomenology is expected to happen in all $\phi^4$-like systems, as well as in the elusive water ice transition across phases VII, VIII and X., Comment: 11 + 4 pages, 12 figures

By means of a combined x-ray diffraction, magnetic susceptibility and specific heat study, we investigate the interplay between orthorhombic distortion and stripe-like antiferromagnetic (AFM) order in the Mott insulator BaCoS$_{2}$ at $T_N=290$ K. The data give evidence of a purely electronic AFM transition with no participation of the lattice. The observation of large thermal fluctuations in the vicinity of $T_N$ and a Schottky anomaly unveils competing ground states within a minute $\sim$1 meV energy range that differ in the orbital and spin configurations of the Co ions. This interpretation suggests that the stripe-like order results from a spontaneous symmetry breaking of the geometrically frustrated pristine tetragonal phase, which offers an ideal playground to study the driving force of multi-orbital Mott transitions without the participation of the lattice., Comment: 5 pages, 4 figures, Supplemental Information

At $T_N\simeq 305~\text{K}$ the layered insulator BaCoS$_2$ transitions to a columnar antiferromagnet that signals non-negligible magnetic frustration despite the relatively high $T_N$, all the more surprising given its quasi two-dimensional structure. Here, we show by combining ab initio and model calculations that the magnetic transition is an order-from-disorder phenomenon, which not only drives the columnar $C_4\to C_2$ symmetry breaking, but also, and more importantly, the inter-layer coherence responsible for the finite N\'eel transition temperature. This uncommon ordering mechanism, actively contributed by orbital degrees of freedom, hints at an abundance of low energy excitations below and, especially, above $T_N$, not in disagreement with experimental evidences, and might as well emerge in other layered correlated compounds showing frustrated magnetism at low temperature., Comment: 12 pages, 6 figures

We present a study of the principal deuterium Hugoniot for pressures up to $150$ GPa, using Machine Learning potentials (MLPs) trained with Quantum Monte Carlo (QMC) energies, forces and pressures. In particular, we adopted a recently proposed workflow based on the combination of Gaussian kernel regression and $\Delta$-learning. By fully taking advantage of this method, we explicitly considered finite-temperature electrons in the dynamics, whose effects are highly relevant for temperatures above $10$ kK. The Hugoniot curve obtained by our MLPs shows a good agreement with the most recent experiments, particularly in the region below 60 GPa. At larger pressures, our Hugoniot curve is slightly more compressible than the one yielded by experiments, whose uncertainties generally increase, however, with pressure. Our work demonstrates that QMC can be successfully combined with $\Delta$-learning to deploy reliable MLPs for complex extended systems across different thermodynamic conditions, by keeping the QMC precision at the computational cost of a mean-field calculation., Comment: 7 + 10 pages; new version with improved QMC dataset. Hugoniot curve and discussion updated accordingly, main physical outcomes unchanged

Water is a key ingredient for life and plays a central role as solvent in many biochemical reactions. However, the intrinsically quantum nature of the hydrogen nucleus, revealing itself in a large variety of physical manifestations, including proton transfer, gives rise to unexpected phenomena whose description is still elusive. Here we study, by an unprecedented combination of state-of-the-art quantum Monte Carlo methods and path-integral molecular dynamics, the structure and hydrogen-bond dynamics of the protonated water hexamer, the fundamental unit for the hydrated proton. We report a remarkably low thermal expansion of the hydrogen bond from zero temperature up to 300 K, owing to the presence of short-Zundel configurations, characterised by proton delocalisation and favoured by the synergy of nuclear quantum effects and thermal activation. The hydrogen bond strength progressively weakens above 300 K, when localised Eigen-like configurations become relevant. Our analysis, supported by the instanton statistics of shuttling protons, reveals that the near-room-temperature range from 250 K to 300 K is a ``sweet spot'' for proton transfer, and thus for many phenomena depending on it, including life., Comment: 25 pages, 5 figures, 23 pages of supplementary information, 14 additional figures in the supplementary information

We propose a new fixed-target experiment for the study of electromagnetic and hard probes of the Quark-Gluon Plasma (QGP) in heavy-ion collisions at the CERN SPS. The experiment aims at performing measurements of the dimuon spectrum from threshold up to the charmonium region, and of hadronic decays of charm and strange hadrons. It is based on a muon spectrometer, which includes a toroidal magnet and six planes of tracking detectors, coupled to a vertex spectrometer, equipped with Si MAPS immersed in a dipole field. High luminosity is an essential requirement for the experiment, with the goal of taking data with 10$^6$ incident ions/s, at collision energies ranging from $\sqrt{s_{\rm NN}} = 6.3$ GeV ($E_{\rm lab}= 20$ A GeV) to top SPS energy ($\sqrt{s_{\rm NN}} = 17.3$ GeV, $E_{\rm lab}= 158$ A GeV). This document presents the physics motivation, the foreseen experimental set-up including integration and radioprotection studies, the current detector choices together with the status of the corresponding R&D, and the outcome of physics performance studies. A preliminary cost evaluation is also carried out., Comment: Letter of Intent submitted to the CERN SPSC