arXiv subjects

Rahul Gupta

Publications and source records attributed to Rahul Gupta.

At least 163 records · Page 9Linked to original sources

arXiv2021

Magnetar Giant Flare Originated GRB 200415A: Transient GeV emission, Time-Resolved $\rm E_p~ -~L_{iso}$ Correlation, and Implications

Giant flares (GFs) are unusual bursts from soft gamma-ray repeaters (SGRs) that release an enormous amount of energy in a fraction of a second. The afterglow emission of these SGR-GFs or GF candidates is a highly beneficial means of discerning their composition, relativistic speed, and emission mechanisms. GRB 200415A is a recent GF candidate observed in a direction coincident with the nearby Sculptor galaxy at 3.5 Mpc. In this work, we searched for transient gamma-ray emission in past observations by Fermi-LAT in the direction of GRB 200415A. These observations confirm that GRB 200415A is observed as a transient GeV source only once. A pure pair-plasma fireball cannot provide the required energy for the interpretation of GeV afterglow emission and a baryonic poor outflow is additionally needed to explain the afterglow emission. A baryonic rich outflow is also viable, as it can explain the variability and observed quasi-thermal spectrum of the prompt emission if dissipation is happening below the photosphere via internal shocks. Using the peak energy ($E_p$) of the time-resolved prompt emission spectra and their fluxes ($F_p$), we found correlation between $E_p$ and $F_p$ or $E_p$ and isotropic luminosity $L_{\rm iso}$ for GRB 200415A. This supports the intrinsic nature of $E_p$-$E_{\rm iso}$ correlation found in SGRs-GFs, hence favoring a baryonic poor outflow. Our results also indicate a different mechanism at work during the initial spike, and that the evolution of the prompt emission spectral properties in this outflow would be intrinsically due to the injection process.

astro-ph.HE↗

arXiv2021

GRB 140102A: Insight into Prompt Spectral Evolution and Early Optical Afterglow Emission

We present and perform a detailed analysis of multi-wavelength observations of \thisgrb, an optical bright GRB with an observed reverse shock (RS) signature. Observations of this GRB were acquired with the BOOTES-4 robotic telescope, the \fermi, and the \swift missions. Time-resolved spectroscopy of the prompt emission shows that changes to the peak energy (\Ep) tracks intensity and the low-energy spectral index seems to follow the intensity for the first episode, whereas this tracking behavior is less clear during the second episode. The fit to the afterglow light curves shows that the early optical afterglow can be described with RS emission and is consistent with the thin shell scenario of the constant ambient medium. The late time afterglow decay is also consistent with the prediction of the external forward shock (FS) model. We determine the properties of the shocks, Lorentz factor, magnetization parameters, and ambient density of \thisgrb, and compare these parameters with another 12 GRBs, consistent with having RS produced by thin shells in an ISM-like medium. The value of the magnetization parameter ($R_{\rm B} \approx 18$) indicates a moderately magnetized baryonic dominant jet composition for \thisgrb. We also report the host galaxy photometric observations of \thisgrb obtained with 10.4m GTC, 3.5m CAHA, and 3.6m DOT telescopes and find the host (photo $z$ = $2.8^{+0.7}_{-0.9}$) to be a high mass, star-forming galaxy with a star formation rate of $20 \pm 10 \msun$ $\rm yr^{-1}$.

astro-ph.HE↗

arXiv2021

Relative $K$-theory via 0-cycles in finite characteristic

Let $R$ be a regular semi-local ring, essentially of finite type over an infinite perfect field of characteristic $p \ge 3$. We show that the cycle class map with modulus from an earlier work of the authors induces a pro-isomorphism between the additive higher Chow groups of relative 0-cycles and the relative $K$-theory of truncated polynomial rings over $R$. This settles the problem of equating 0-cycles with modulus and relative $K$-theory of such rings via the cycle class map.

math.AG↗

arXiv2021

Progenitor mass constraints for the type Ib intermediate-luminosity SN 2015ap and the highly extinguished SN 2016bau

Photometric and spectroscopic analyses of the intermediate-luminosity Type Ib supernova (SN) 2015ap and of the heavily reddened Type Ib SN~2016bau are discussed. Photometric properties of the two SNe, such as colour evolution, bolometric luminosity, photospheric radius, temperature, and velocity evolution, are also constrained. The ejecta mass, synthesised nickel mass, and kinetic energy of the ejecta are calculated from their light-curve analysis. We also model and compare the spectra of SN~2015ap and SN~2016bau at various stages of their evolution. The P~Cygni profiles of various lines present in the spectra are used to determine the velocity evolution of the ejecta. To account for the observed photometric and spectroscopic properties of the two SNe, we have computed 12\,$M_\odot$ zero-age main sequence (ZAMS) star models and evolved them until the onset of core collapse using the publicly available stellar-evolution code {\tt MESA}. Synthetic explosions were produced using the public version of {\tt STELLA} and another publicly available code, {\tt SNEC}, utilising the {\tt MESA} models. {\tt SNEC} and {\tt STELLA} provide various observable properties such as the bolometric luminosity and velocity evolution. The parameters produced by {\tt SNEC}/{\tt STELLA} and our observations show close agreement with each other, thus supporting a 12\,$M_\odot$ ZAMS star as the possible progenitor for SN~2015ap, while the progenitor of SN~2016bau is slightly less massive, being close to the boundary between SN and non-SN as the final product.

astro-ph.SR↗

arXiv2021

Giant spin-orbit torque efficiency in all-epitaxial heterostructures

A large anti-damping spin-obit torque (SOT) efficiency in magnetic heterostructures is a prerequisite to realize energy efficient spin torque based magnetic memories and logic devices. The efficiency can be characterized in terms of the spin-orbit fields generated by anti-damping torques when an electric current is passed through the non-magnetic layer. We report a giant spin-orbit field of 48.96 (27.50) mT at an applied current density of 1 MAcm-2 in beta-W interfaced Co60Fe40 (Ni81Fe19)/TiN epitaxial structures due to an anti-damping like torque, which results in a magnetization auto-oscillation current density as low as 1.68(3.27) MAcm-2. The spin-orbit field value increases with decrease of beta-W layer thickness, which affirms that epitaxial surface states are responsible for the extraordinary large efficiency. SOT induced energy efficient in-plane magnetization switching in large 20x100 um2 structures has been demonstrated by Kerr microscopy and the findings are supported by results from micromagnetic simulations. The observed giant SOT efficiencies in the studied all-epitaxial heterostructures are comparable to values reported for topological insulators. These results confirm that by utilizing epitaxial material combinations an extraordinary large SOT efficiency can be achieved using semiconducting industry compatible 5d heavy metals, which provides immediate solutions for the realization of energy efficient spin-logic devices.

cond-mat.mtrl-sci↗

arXiv2021

Ramified class field theory and duality over finite fields

We prove a duality theorem for the $p$-adic etale motivic cohomology of a variety $U$ which is the complement of a divisor on a smooth projective variety over $\F_p$. This extends the duality theorems of Milne and Jannsen-Saito-Zhao. The duality introduces a filtration on $H^1_{\etl}(U, {\Q}/{\Z})$. We identify this filtration to the classically known Matsuda filtration when the reduced part of the divisor is smooth. We prove a reciprocity theorem for the idele class groups with modulus introduced by Kerz-Zhao and Rulling-Saito. As an application, we derive the failure of Nisnevich descent for Chow groups with modulus.

math.AG↗

arXiv2021

ADePT: Auto-encoder based Differentially Private Text Transformation

Privacy is an important concern when building statistical models on data containing personal information. Differential privacy offers a strong definition of privacy and can be used to solve several privacy concerns (Dwork et al., 2014). Multiple solutions have been proposed for the differentially-private transformation of datasets containing sensitive information. However, such transformation algorithms offer poor utility in Natural Language Processing (NLP) tasks due to noise added in the process. In this paper, we address this issue by providing a utility-preserving differentially private text transformation algorithm using auto-encoders. Our algorithm transforms text to offer robustness against attacks and produces transformations with high semantic quality that perform well on downstream NLP tasks. We prove the theoretical privacy guarantee of our algorithm and assess its privacy leakage under Membership Inference Attacks(MIA) (Shokri et al., 2017) on models trained with transformed data. Our results show that the proposed model performs better against MIA attacks while offering lower to no degradation in the utility of the underlying transformation process compared to existing baselines.

cs.CR↗

arXiv2021

ProtoDA: Efficient Transfer Learning for Few-Shot Intent Classification

Practical sequence classification tasks in natural language processing often suffer from low training data availability for target classes. Recent works towards mitigating this problem have focused on transfer learning using embeddings pre-trained on often unrelated tasks, for instance, language modeling. We adopt an alternative approach by transfer learning on an ensemble of related tasks using prototypical networks under the meta-learning paradigm. Using intent classification as a case study, we demonstrate that increasing variability in training tasks can significantly improve classification performance. Further, we apply data augmentation in conjunction with meta-learning to reduce sampling bias. We make use of a conditional generator for data augmentation that is trained directly using the meta-learning objective and simultaneously with prototypical networks, hence ensuring that data augmentation is customized to the task. We explore augmentation in the sentence embedding space as well as prototypical embedding space. Combining meta-learning with augmentation provides upto 6.49% and 8.53% relative F1-score improvements over the best performing systems in the 5-shot and 10-shot learning, respectively.

cs.CL↗

arXiv2021

BOLD: Dataset and Metrics for Measuring Biases in Open-Ended Language Generation

Recent advances in deep learning techniques have enabled machines to generate cohesive open-ended text when prompted with a sequence of words as context. While these models now empower many downstream applications from conversation bots to automatic storytelling, they have been shown to generate texts that exhibit social biases. To systematically study and benchmark social biases in open-ended language generation, we introduce the Bias in Open-Ended Language Generation Dataset (BOLD), a large-scale dataset that consists of 23,679 English text generation prompts for bias benchmarking across five domains: profession, gender, race, religion, and political ideology. We also propose new automated metrics for toxicity, psycholinguistic norms, and text gender polarity to measure social biases in open-ended text generation from multiple angles. An examination of text generated from three popular language models reveals that the majority of these models exhibit a larger social bias than human-written Wikipedia text across all domains. With these results we highlight the need to benchmark biases in open-ended language generation and caution users of language generation models on downstream tasks to be cognizant of these embedded prejudices.

cs.CL↗

arXiv2021

Co$_2$FeAl full Heusler compound based spintronic terahertz emitter

To achieve a large terahertz (THz) amplitude from a spintronic THz emitter (STE), materials with 100\% spin polarisation such as Co-based Heusler compounds as the ferromagnetic layer are required. However, these compounds are known to loose their half-metallicity in the ultrathin film regime, as it is difficult to achieve L2$_1$ ordering, which has become a bottleneck for the film growth. Here, the successful deposition using room temperature DC sputtering of the L2$_1$ and B2 ordered phases of the Co$_2$FeAl full Heusler compound is reported. Co$_2$FeAl is used as ferromagnetic layer together with highly orientated Pt as non-ferromagnetic layer in the Co$_2$FeAl/Pt STE, where an MgO(10 nm) seed layer plays an important role to achieve the L2$_1$ and B2 ordering of Co$_2$FeAl. The generation of THz radiation in the CFA/Pt STE is presented, which has a bandwidth in the range of 0.1-4 THz. The THz electric field amplitude is optimized with respect to thickness, orientation, and growth parameters using a thickness dependent model considering the optically induced spin current, superdiffusive spin current, inverse spin Hall effect and the attenuation of THz radiation in the layers. This study, based on the full Heusler Co$_2$FeAl compound opens up a plethora possibilities in STE research involving full Heusler compounds.

cond-mat.mtrl-sci↗

arXiv2021

SN 2020ank: a bright and fast-evolving H-deficient superluminous supernova

We investigate the observational properties of a hydrogen-deficient superluminous supernova (SLSN) SN 2020ank (at z = 0.2485), with the help of early phase observations carried out between $-$21 and +52 d since $g$-band maximum. Photometrically, SN 2020ank is one of the brightest SLSN ($M_{g,peak}$ $\sim$ $-$21.84 $\pm$ 0.10 mag), having fast pre-peak rising and post-peak decaying rates. The bolometric light curve of SN 2020ank exhibits a higher peak luminosity ($L_{max}$) of $\sim$(3.9 $\pm$ 0.7) $\times$ 10$^{44}$ erg s$^{-1}$ and appears to be symmetric around the peak with $L^{rise}_{max}$/e $\approx$ $L^{fall}_{max}$/e $\approx$ 15 d. The semi-analytical light-curve modelling using the MINIM code suggests a spin down millisecond magnetar with $P_i$ $\sim$2.2 $\pm$ 0.5 ms and $B$ $\sim$(2.9 $\pm$ 0.1) $\times$ $10^{14}$ G as a possible powering source for SN 2020ank. The possible magnetar origin and excess ultraviolet flux at early epochs indicate a central-engine based powering source for SN 2020ank. Near-peak spectra of SN 2020ank are enriched with the W-shaped O II features but with the weaker signatures of C II and Fe III. Using the estimated rise time of $\sim$27.9 d and the photospheric velocity of $\sim$12050 km s$^{-1}$, we constrain the ejecta mass to $\sim$7.2 $M_{\odot}$ and the kinetic energy of $\sim$6.3 $\times$ 10$^{51}$ erg. The near-peak spectrum of SN 2020ank exhibits a close spectral resemblance with that of fast-evolving SN 2010gx. The absorption features of SN 2020ank are blueshifted compared to Gaia16apd, suggesting a higher expansion velocity. The spectral similarity with SN 2010gx and comparatively faster spectral evolution than PTF12dam (a slow-evolving SLSN) indicate the fast-evolving behavior of SN 2020ank.

astro-ph.HE↗

arXiv2020

Evaluating the Effectiveness of Efficient Neural Architecture Search for Sentence-Pair Tasks

Neural Architecture Search (NAS) methods, which automatically learn entire neural model or individual neural cell architectures, have recently achieved competitive or state-of-the-art (SOTA) performance on variety of natural language processing and computer vision tasks, including language modeling, natural language inference, and image classification. In this work, we explore the applicability of a SOTA NAS algorithm, Efficient Neural Architecture Search (ENAS) (Pham et al., 2018) to two sentence pair tasks, paraphrase detection and semantic textual similarity. We use ENAS to perform a micro-level search and learn a task-optimized RNN cell architecture as a drop-in replacement for an LSTM. We explore the effectiveness of ENAS through experiments on three datasets (MRPC, SICK, STS-B), with two different models (ESIM, BiLSTM-Max), and two sets of embeddings (Glove, BERT). In contrast to prior work applying ENAS to NLP tasks, our results are mixed -- we find that ENAS architectures sometimes, but not always, outperform LSTMs and perform similarly to random architecture search.

cs.CL↗

arXiv2020

Phase Transition Behavior in Knowledge Compilation

The study of phase transition behaviour in SAT has led to deeper understanding and algorithmic improvements of modern SAT solvers. Motivated by these prior studies of phase transitions in SAT, we seek to study the behaviour of size and compile-time behaviour for random k-CNF formulas in the context of knowledge compilation. We perform a rigorous empirical study and analysis of the size and runtime behavior for different knowledge compilation forms (and their corresponding compilation algorithms): d-DNNFs, SDDs and OBDDs across multiple tools and compilation algorithms. We employ instances generated from the random k-CNF model with varying generation parameters to empirically reason about the expected and median behavior of size and compilation-time for these languages. Our work is similar in spirit to the early work in CSP community on phase transition behavior in SAT/CSP. In a similar spirit, we identify the interesting behavior with respect to different parameters: clause density and solution density, a novel control parameter that we identify for the study of phase transition behavior in the context of knowledge compilation. Furthermore, we summarize our empirical study in terms of two concrete conjectures; a rigorous study of these conjectures will possibly require new theoretical tools.

cs.AI↗

arXiv2020

Automatic Discovery of Novel Intents & Domains from Text Utterances

One of the primary tasks in Natural Language Understanding (NLU) is to recognize the intents as well as domains of users' spoken and written language utterances. Most existing research formulates this as a supervised classification problem with a closed-world assumption, i.e. the domains or intents to be identified are pre-defined or known beforehand. Real-world applications however increasingly encounter dynamic, rapidly evolving environments with newly emerging intents and domains, about which no information is known during model training. We propose a novel framework, ADVIN, to automatically discover novel domains and intents from large volumes of unlabeled data. We first employ an open classification model to identify all utterances potentially consisting of a novel intent. Next, we build a knowledge transfer component with a pairwise margin loss function. It learns discriminative deep features to group together utterances and discover multiple latent intent categories within them in an unsupervised manner. We finally hierarchically link mutually related intents into domains, forming an intent-domain taxonomy. ADVIN significantly outperforms baselines on three benchmark datasets, and real user utterances from a commercial voice-powered agent.

cs.CL↗

arXiv2020

Peculiar prompt emission and afterglow in H.E.S.S. detected GRB 190829A

We present the results of a detailed investigation of the prompt and afterglow emission in the HESS detected GRB 190829A. Swift and Fermi observations of the prompt phase of this GRB reveal two isolated sub-bursts or episodes, separated by a quiescent phase. The energetic and the spectral properties of the first episode are in stark contrast to the second. The first episode, which has a higher spectral peak of $\sim 120\:\text{keV}$ and a low isotropic energy $\sim 10^{50}\:\text{erg}$ is an outlier to the Amati correlation and marginally satisfies the Yonetoku correlation. However, the energetically dominant second episode has lower peak energy and is consistent with the above correlations. We compared this GRB to other low luminosity GRBs (LLGRBs). Prompt emission of LLGRBs also indicates a relativistic shock breakout origin of the radiation. For GRB 190829A, some of the properties of a shock breakout origin are satisfied. However, the absence of an accompanying thermal component and energy above the shock breakout critical limit precludes a shock breakout origin. In the afterglow, an unusual long-lasting late time flare of duration $\sim 10^4\:\text{s}$ is observed. We also analyzed the late-time \fermi-LAT emission that encapsulates the H.E.S.S. detection. Some of the LAT photons are likely to be associated with the source. All the above observational facts suggest GRB 190829A is a peculiar low luminosity GRB that is not powered by a shock breakout, and with an unusual rebrightening due to a patchy emission or a refreshed shock during the afterglow. Furthermore, our results show that TeV energy photons seem common in both high luminosity GRBs and LLGRBs.

astro-ph.HE↗

arXiv2020

Towards classification parity across cohorts

Recently, there has been a lot of interest in ensuring algorithmic fairness in machine learning where the central question is how to prevent sensitive information (e.g. knowledge about the ethnic group of an individual) from adding "unfair" bias to a learning algorithm (Feldman et al. (2015), Zemel et al. (2013)). This has led to several debiasing algorithms on word embeddings (Qian et al. (2019) , Bolukbasi et al. (2016)), coreference resolution (Zhao et al. (2018a)), semantic role labeling (Zhao et al. (2017)), etc. Most of these existing work deals with explicit sensitive features such as gender, occupations or race which doesn't work with data where such features are not captured due to privacy concerns. In this research work, we aim to achieve classification parity across explicit as well as implicit sensitive features. We define explicit cohorts as groups of people based on explicit sensitive attributes provided in the data (age, gender, race) whereas implicit cohorts are defined as groups of people with similar language usage. We obtain implicit cohorts by clustering embeddings of each individual trained on the language generated by them using a language model. We achieve two primary objectives in this work : [1.] We experimented and discovered classification performance differences across cohorts based on implicit and explicit features , [2] We improved classification parity by introducing modification to the loss function aimed to minimize the range of model performances across cohorts.

cs.LG↗

arXiv2020

Zero-cycles with modulus and relative $K$-theory

We construct a cycle class map from the higher Chow groups of 0-cycles to the relative $K$-theory of a modulus pair. We show that this induces a pro-isomorphism between the additive higher Chow groups of relative 0-cycles and relative $K$-theory of truncated polynomial rings over a regular semi-local ring, essentially of finite type over a characteristic zero field.

math.AG↗

arXiv2020

Joint Multi-Dimensional Model for Global and Time-Series Annotations

Crowdsourcing is a popular approach to collect annotations for unlabeled data instances. It involves collecting a large number of annotations from several, often naive untrained annotators for each data instance which are then combined to estimate the ground truth. Further, annotations for constructs such as affect are often multi-dimensional with annotators rating multiple dimensions, such as valence and arousal, for each instance. Most annotation fusion schemes however ignore this aspect and model each dimension separately. In this work we address this by proposing a generative model for multi-dimensional annotation fusion, which models the dimensions jointly leading to more accurate ground truth estimates. The model we propose is applicable to both global and time series annotation fusion problems and treats the ground truth as a latent variable distorted by the annotators. The model parameters are estimated using the Expectation-Maximization algorithm and we evaluate its performance using synthetic data and real emotion corpora as well as on an artificial task with human annotations

cs.LG↗