arXiv subjects

Vivek Kumar

Publications and source records attributed to Vivek Kumar.

At least 55 records · Page 3Linked to original sources

arXiv2024

Unlocking LLMs: Addressing Scarce Data and Bias Challenges in Mental Health

Large language models (LLMs) have shown promising capabilities in healthcare analysis but face several challenges like hallucinations, parroting, and bias manifestation. These challenges are exacerbated in complex, sensitive, and low-resource domains. Therefore, in this work we introduce IC-AnnoMI, an expert-annotated motivational interviewing (MI) dataset built upon AnnoMI by generating in-context conversational dialogues leveraging LLMs, particularly ChatGPT. IC-AnnoMI employs targeted prompts accurately engineered through cues and tailored information, taking into account therapy style (empathy, reflection), contextual relevance, and false semantic change. Subsequently, the dialogues are annotated by experts, strictly adhering to the Motivational Interviewing Skills Code (MISC), focusing on both the psychological and linguistic dimensions of MI dialogues. We comprehensively evaluate the IC-AnnoMI dataset and ChatGPT's emotional reasoning ability and understanding of domain intricacies by modeling novel classification tasks employing several classical machine learning and current state-of-the-art transformer approaches. Finally, we discuss the effects of progressive prompting strategies and the impact of augmented data in mitigating the biases manifested in IC-AnnoM. Our contributions provide the MI community with not only a comprehensive dataset but also valuable insights for using LLMs in empathetic text generation for conversational therapy in supervised settings.

cs.CL↗

arXiv2024

Terahertz Microscopy Through Complex Media

Manipulating broadband fields in scattering media is a modern challenge across photonics and other wave domains. Recent studies have shown that complex propagation in scattering media can be harnessed to manipulate broadband light wave packets in space-time for focusing, imaging, and computing applications. Interestingly, while many proposed methodologies operate on intensity-based assessment of scattered fields, often in the spectral domain, from a pure transmission-function perspective, scattering operates as a linear field-level combinatory process, i.e., the superposition of transformation of unit excitations. As a result, we recently demonstrated that gaining experimental access to instantaneous scattered fields, as available through time-domain spectroscopy in the terahertz spectral range, in conjunction with sparse light excitation typical of ghost imaging, provides a key advantage in enabling the functionalisation of scattering, exposing a novel modelling paradigm. In this paper, we provide experimental proof of reconstructing 1-dimensional object features through a scattering medium using a fully broadband time-domain terahertz approach.

physics.optics↗

arXiv2024

Fourier Synthetic Aperture-based Time-resolved Terahertz Imaging

Terahertz microscopy has attracted attention owing to distinctive characteristics of the THz frequency region, particularly non-ionizing photon energy, spectral fingerprint, and transparency to most nonpolar materials. Nevertheless, the well-known Rayleigh diffraction limit imposed on THz waves commonly constrains the resultant imaging resolution to values beyond the millimeter scale, consequently limiting the applicability in numerous emerging applications for chemical sensing and complex media imaging. In this theoretical and numerical work, we address this challenge by introducing a new imaging approach, based on acquiring high-spatial frequencies by adapting the Fourier synthetic aperture approach to the terahertz spectral range, thus surpassing the diffraction-limited resolution. Our methodology combines multi-angle terahertz pulsed illumination with time-resolved field measurements, as enabled by the state-of-the-art time-domain spectroscopy technique. We demonstrate the potential of the approach for hyperspectral terahertz imaging of semi-transparent samples and show that the technique can reconstruct spatial and temporal features of complex inhomogeneous samples with subwavelength resolution.

physics.optics↗

arXiv2024

Central limit theorem and moderate deviation principle for the stochastic generalized Burgers-Huxley equation with multiplicative noise

In this work, we investigate the Central Limit Theorem (CLT) and Moderate Deviation Principle (MDP) for the stochastic generalized Burgers-Huxley (SGBH) equation with multiplicative Gaussian noise. The SGBH equation is a diffusion-convection-reaction type equation which consists a nonlinearity of polynomial order, and we take into account of an infinite-dimensional noise having a coefficient that has linear growth. We first prove the CLT which allows us to establish the convergence of the distribution of the solution to a re-scaled SGBH equation to a desired distribution function. Furthermore, we extend our asymptotic analysis by investigating the MDP for the SGBH equation. Using the weak convergence method, we establish the MDP and derive the corresponding rate function.

math.PR↗

arXiv2024

Surface phonons and possible structural phase transition in a topological semimetal PbTaSe2

Topological insulators are a novel class of quantum materials characterized by protected gapless surface or edge states but insulating bulk states which is due to presence of spin-orbit interactions and time-reversal symmetry. Such an intriguing surface and bulk topology manifests itself in coupling with lattice dynamics due to electron-phonon scattering. Here we report an in-depth investigation of a topological nodal line semimetal PbTaSe2 via temperature, polarization dependent Raman spectroscopy and temperature dependent single crystal X-ray diffraction (SC-XRD) measurements. Our analysis shows signature of electron-phonon coupling as reflected in the Fano asymmetry in line shape of M1-M4 modes and anomalous temperature variation of line-width of P3-P4 modes. Further polarization dependent phonon symmetry changes at different temperature (6K and 300K), discontinuities in bulk phonon dynamics for P2-P5 modes and disappearance of phonon modes i.e., M1-M5, on decreasing temperatures indicates towards a thermally induced structural phase transition which is also supported by the SC-XRD results. Hence based on our findings we propose that M1-M4 modes are surface phonon modes, the material undergoes a thermally induced structural phase transition from alpha to beta phase at T ~ 150 K or is in close proximity to the beta phase and another transition below T(CDW+beta) ~ 100K which is possibly due to the interplay of remanent completely commensurate charge density wave (CCDW) of 1H-TaSe2 and beta phase.

cond-mat.str-el↗

arXiv2024

Ab-initio investigations of novel potential all-d metal Heusler alloys Co$_2$MnNb

In this study, we employ the Wien2k code to conduct ab-initio study of a novel potential all-d-metal Heusler alloy Co$_2$MnNb. The analysis utilizes the comparison of local spin density approximations (LDA) with Perdew-Burke-Ernzerh parameterized Generalized Gradient Approximation (PBE-GGA) for structural optimization while modified Becke-Jones potential (mBJ) exchange-correlation potentials to examine various characteristic properties of the alloy under study. Employing Birch-Murnaghan equation of state, we construct the energy-versus-volume curve, facilitating the determination of stable phases and structural parameters of the investigated alloys. Structural optimization in both non-magnetic (NM) and spin-polarized (FM) states reveals the stability of the alloy in the FM state. The compound exhibits metallic behavior in bulk, with notable anisotropic semiconducting behavior for down spin while pure metallic behavior for up spin electrons. Partial density of states of each element of the composition is also analysed to compare their respective contribution towards the observed band structure. The anisotropic behavior of Co$_2$MnNb for a specific spin state could be of importance in future spintronic and other thin films device applications.

cond-mat.mtrl-sci↗

arXiv2024

PhilHumans: Benchmarking Machine Learning for Personal Health

The use of machine learning in Healthcare has the potential to improve patient outcomes as well as broaden the reach and affordability of Healthcare. The history of other application areas indicates that strong benchmarks are essential for the development of intelligent systems. We present Personal Health Interfaces Leveraging HUman-MAchine Natural interactions (PhilHumans), a holistic suite of benchmarks for machine learning across different Healthcare settings - talk therapy, diet coaching, emergency care, intensive care, obstetric sonography - as well as different learning settings, such as action anticipation, timeseries modeling, insight mining, language modeling, computer vision, reinforcement learning and program synthesis

cs.LG↗

arXiv2024

Ask the experts: sourcing high-quality datasets for nutritional counselling through Human-AI collaboration

Large Language Models (LLMs), with their flexible generation abilities, can be powerful data sources in domains with few or no available corpora. However, problems like hallucinations and biases limit such applications. In this case study, we pick nutrition counselling, a domain lacking any public resource, and show that high-quality datasets can be gathered by combining LLMs, crowd-workers and nutrition experts. We first crowd-source and cluster a novel dataset of diet-related issues, then work with experts to prompt ChatGPT into producing related supportive text. Finally, we let the experts evaluate the safety of the generated text. We release HAI-coaching, the first expert-annotated nutrition counselling dataset containing ~2.4K dietary struggles from crowd workers, and ~97K related supportive texts generated by ChatGPT. Extensive analysis shows that ChatGPT while producing highly fluent and human-like text, also manifests harmful behaviours, especially in sensitive topics like mental health, making it unsuitable for unsupervised use.

cs.CL↗

arXiv2023

Towards a comprehensive understanding of the low energy luminescence peak in 2D materials

An intense low-energy broad luminescence peak (L-peak) is usually observed in 2D transition metal dichalcogenides (TMDs) at low temperatures. L-peak has earlier been attributed to bound excitons, but its origins are widely debated with direct consequences on optoelectronic properties. To decouple the contributions of physisorbed and chemisorbed oxygen, organic adsorbates, and strain on L-peak, we measured a series of monolayer (ML) MoS2 samples (mechanically exfoliated (ME), synthesized by oxygen-assisted chemical vapour deposition (O-CVD), hexagonal boron nitride (hBN) covered and hBN encapsulated). Emergence of L-peak below 150 K and saturation of photoluminescence (PL) intensity with laser power confirm bound nature of L-peak. Anomalously at room temperature, O-CVD samples show high A-exciton PL (c.f. ME), but reduced PL at low temperatures, which is attributed to strain-induced direct-to-indirect bandgap change in low defect O-CVD MoS2. Further, L-peak redshifts dramatically ~ 130 meV for O-CVD samples (c.f. ME). These observations are fully consistent with our predictions from density functional theory (DFT) calculations, considering effects of both strain and defects, and supported by Raman spectroscopy. In ME samples, charged oxygen adatoms are identified as thermodynamically favourable defects which can create in-gap states, and contribute to the L-peak. The useful effect of hBN is found to originate from reduction of charged oxygen adatoms and hydrocarbon complexes. This combined experimental-theoretical study allows an enriched understanding of L-peak and beneficial impact of hBN, and motivates collective studies of strain and defects with direct impact on optoelectronics and quantum technologies.

cond-mat.mtrl-sci↗

arXiv2023

Regularity and numerical approximation of fractional elliptic differential equations on compact metric graphs

The fractional differential equation $L^βu = f$ posed on a compact metric graph is considered, where $β>0$ and $L = κ^2 - \nabla(a\nabla)$ is a second-order elliptic operator equipped with certain vertex conditions and sufficiently smooth and positive coefficients $κ, a$. We demonstrate the existence of a unique solution for a general class of vertex conditions and derive the regularity of the solution in the specific case of Kirchhoff vertex conditions. These results are extended to the stochastic setting when $f$ is replaced by Gaussian white noise. For the deterministic and stochastic settings under generalized Kirchhoff vertex conditions, we propose a numerical solution based on a finite element approximation combined with a rational approximation of the fractional power $L^{-β}$. For the resulting approximation, the strong error is analyzed in the deterministic case, and the strong mean squared error as well as the $L_2(Γ\times Γ)$-error of the covariance function of the solution are analyzed in the stochastic setting. Explicit rates of convergences are derived for all cases. Numerical experiments for ${L = κ^2 - Δ, κ>0}$ are performed to illustrate the results.

math.NA↗

arXiv2023

SPAE: Semantic Pyramid AutoEncoder for Multimodal Generation with Frozen LLMs

In this work, we introduce Semantic Pyramid AutoEncoder (SPAE) for enabling frozen LLMs to perform both understanding and generation tasks involving non-linguistic modalities such as images or videos. SPAE converts between raw pixels and interpretable lexical tokens (or words) extracted from the LLM's vocabulary. The resulting tokens capture both the semantic meaning and the fine-grained details needed for visual reconstruction, effectively translating the visual content into a language comprehensible to the LLM, and empowering it to perform a wide array of multimodal tasks. Our approach is validated through in-context learning experiments with frozen PaLM 2 and GPT 3.5 on a diverse set of image understanding and generation tasks. Our method marks the first successful attempt to enable a frozen LLM to generate image content while surpassing state-of-the-art performance in image understanding tasks, under the same setting, by over 25%.

cs.CV↗

arXiv2023

Implementation of a level-set based volume penalization method for solving fluid flows around bluff bodies in OpenFOAM

A volume penalization-based immersed boundary technique is developed and thoroughly validated for fluid flow problems, specifically flow over bluff bodies. The proposed algorithm has been implemented in an Open Source Field Operation and Manipulation (OpenFOAM). For capturing the fluid-solid interface more accurately, the grid is refined near the solid surface using topoSetDict and refineMeshDict utilities in OpenFOAM. In order to avoid any numerical oscillation, the present volume penalization method (VPM) is integrated with a signed distance function, which is also referred to as a level-set function. Benchmark problems, such as flows around a cylinder and a sphere, are considered and thoroughly validated with the results available in the literature. For the flow over a stationary cylinder, the Reynolds number is varied so that it covers from a steady 2D (two-dimensional) flow to an unsteady 3D (three-dimensional) flow. The capability of the present solver has been further verified by considering the flow past a vibrating cylinder in the cross-stream direction. In addition, a flow over a sphere, which is inherently three-dimensional due to its geometrical shape, is validated in both steady and unsteady regimes. The results obtained by the present VPM show good agreement with those obtained by a body-fitted grid using the same numerical scheme as that of the VPM, and also with those reported in the literature. The present results indicate that the VPM-based immersed boundary technique can be widely applicable to scientific and engineering problems involving flow past stationary and moving bluff bodies of arbitrary geometry.

physics.flu-dyn↗

arXiv2023

VISU at WASSA 2023 Shared Task: Detecting Emotions in Reaction to News Stories Leveraging BERT and Stacked Embeddings

Our system, VISU, participated in the WASSA 2023 Shared Task (3) of Emotion Classification from essays written in reaction to news articles. Emotion detection from complex dialogues is challenging and often requires context/domain understanding. Therefore in this research, we have focused on developing deep learning (DL) models using the combination of word embedding representations with tailored prepossessing strategies to capture the nuances of emotions expressed. Our experiments used static and contextual embeddings (individual and stacked) with Bidirectional Long short-term memory (BiLSTM) and Transformer based models. We occupied rank tenth in the emotion detection task by scoring a Macro F1-Score of 0.2717, validating the efficacy of our implemented approaches for small and imbalanced datasets with mixed categories of target emotions.

cs.CL↗

arXiv2023

MABe22: A Multi-Species Multi-Task Benchmark for Learned Representations of Behavior

We introduce MABe22, a large-scale, multi-agent video and trajectory benchmark to assess the quality of learned behavior representations. This dataset is collected from a variety of biology experiments, and includes triplets of interacting mice (4.7 million frames video+pose tracking data, 10 million frames pose only), symbiotic beetle-ant interactions (10 million frames video data), and groups of interacting flies (4.4 million frames of pose tracking data). Accompanying these data, we introduce a panel of real-life downstream analysis tasks to assess the quality of learned representations by evaluating how well they preserve information about the experimental conditions (e.g. strain, time of day, optogenetic stimulation) and animal behavior. We test multiple state-of-the-art self-supervised video and trajectory representation learning methods to demonstrate the use of our benchmark, revealing that methods developed using human action datasets do not fully translate to animal datasets. We hope that our benchmark and dataset encourage a broader exploration of behavior representation learning methods across species and settings.

cs.LG↗

arXiv2023

Optimal Entanglement Distillation Policies for Quantum Switches

In an entanglement distribution network, the function of a quantum switch is to generate elementary entanglement with its clients followed by entanglement swapping to distribute end-to-end entanglement of sufficiently high fidelity between clients. The threshold on entanglement fidelity is any quality-of-service requirement specified by the clients as dictated by the application they run on the network. We consider a discrete-time model for a quantum switch that attempts generation of fresh elementary entanglement with clients in each time step in the form of maximally entangled qubit pairs, or Bell pairs, which succeed probabilistically; the successfully generated Bell pairs are stored in noisy quantum memories until they can be swapped. We focus on establishing the value of entanglement distillation of the stored Bell pairs prior to entanglement swapping in presence of their inevitable aging, i.e., decoherence: For a simple instance of a switch with two clients, exponential decay of entanglement fidelity, and a well-known probabilistic but heralded two-to-one distillation protocol, given a threshold end-to-end entanglement fidelity, we use the Markov Decision Processes framework to identify the optimal action policy - to wait, to distill, or to swap that maximizes throughput. We compare the switch's performance under the optimal distillation-enabled policy with that excluding distillation. Simulations of the two policies demonstrate the improvements that are possible in principle via optimal use of distillation with respect to average throughput, average fidelity, and jitter of end-to-end entanglement, as functions of fidelity threshold. Our model thus helps capture the role of entanglement distillation in mitigating the effects of decoherence in a quantum switch in an entanglement distribution network, adding to the growing literature on quantum switches.

quant-ph↗

arXiv2023

Well-posedness and uniform large deviation principle for stochastic Burgers-Huxley equation perturbed by a multiplicative noise

In this work, we focus on the global solvability and uniform large deviations for the solutions of stochastic generalized Burgers-Huxley (SGBH) equation perturbed by a small multiplicative white in time and colored in space noise. The SGBH equation has the nonlinearity of polynomial order and noise considered in this work is infinite dimensional with a coefficient having linear growth. First, we prove the existence of a \textsl{unique local mild solution} in the sense of Walsh to SGBH equation with the help of a truncation argument and contraction mapping principle. Then the global solvability results are established by using uniform bounds of the local mild solution, stopping time arguments, tightness properties and Skorokhod's representation theorem. By using the uniform Laplace principle, we obtain the \textsl{large deviation principle} (LDP) for the law of solutions to SGBH equation by using variational representation methods. Further, we derive the \textsl{uniform large deviation principle} (ULDP) for the law of solutions in two different topologies by using a weak convergence method. First, in the $\mathrm{C}([0, T ];\mathrm{L}^p([0,1])) $ topology where the uniformity is over $\mathrm{L}^p([0,1])$-bounded sets of initial conditions, and secondly in the $\mathrm{C} ([0, T ] \times[0,1])$ topology with uniformity being over bounded subsets in the $\mathrm{C}([0,1])$-norm. Finally, we consider SGBH equation perturbed by a space-time white noise with bounded noise coefficient and establish the ULDP for the laws of solutions. The results obtained in this work hold true for stochastic Burgers' as well as Burgers-Huxley equations.

math.PR↗

arXiv2023

Low-temperature antiferromagnetic order in orthorhombic CePdAl$_{3}$

We report the magnetization, ac susceptibility, and specific heat of optically float-zoned single crystals of CePdAl$_{3}$. In comparison to the properties of polycrystalline CePdAl$_{3}$ reported in the literature, which displays a tetragonal crystal structure and no long-range magnetic order, our single crystals exhibit an orthorhombic structure ($Cmcm$) and order antiferromagnetically below a Néel temperature $T_{\rm N}$ = 5.6 K. The specific heat at zero-field shows a clear $λ$-type anomaly with a broad shoulder at $T_{\rm N}$. A conservative estimate of the Sommerfeld coefficient of the electronic specific heat, $γ= 121~\mathrm{mJ~K^{-2}~mol^{-1}}$, indicates a moderately enhanced heavy-fermion ground state. A twin microstructure evolves in the family of planes spanned by the basal plane lattice vectors $a_{\rm o}$ and $c_{\rm o}$, with the magnetic hard axis $b_{\rm o}$ common to all twins. The antiferromagnetic state is characterized by a strong magnetic anisotropy and a spin-flop transition induced under magnetic field along the easy direction, resulting in a complex magnetic phase diagram. Taken together our results reveal a high sensitivity of the magnetic and electronic properties of CePdAl$_{3}$ to its structural modifications.

cond-mat.str-el↗

arXiv2022

Grain boundary and defects assisted thermal conductivity of nano-crystalline Gd2Ti2O7 Pyrochlore

The thermal conductivity study on the pyrochlore structured ceramics is important for utilization of these materials as an inert matrix fuel, electrolytes for oxide fuel cell and thermal barrier coating. The impact of porosity, structural defects and boundary scattering on thermal properties of the nanocrystalline Gd2Ti2O7 fabricated by spark plasma sintering followed by high energy ball milling has been investigated. The thermal conductivity has been measured in the temperature range from room temperature to 900 C using laser flash apparatus (LFA), and results were discussed by including several phenomena contributing to the thermal transmission in nc-pyrochlore in support of the scattering mechanism. A systematic decrease in thermal conductivity with reducing grain size can be understood by the phonon-pore scattering phenomenon. Furthermore, dominate of the boundary scattering on the thermal transmission with the increasing temperature is also observed. The experimentally determined values of thermal conductivity are compatible with the thermal insulation requirements for thermal barrier coating (TBC) applications

cond-mat.mtrl-sci↗