arXiv subjects

Rahul Gupta

Publications and source records attributed to Rahul Gupta.

At least 181 records · Page 10Linked to original sources

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↗

arXiv2020

Damping-like Torque in Monolayer 1T-TaS$_2$

A damping-like spin orbit torque (SOT) is a prerequisite for ultralow power spin logic devices. Here, we report on the damping-like SOT in just one monolayer of the conducting transition metal dichalcogenide (TMD) TaS$_2$ interfaced with a NiFe (Py) ferromagnetic layer. The charge-spin conversion efficiency is found to be 0.25$\pm$0.03 and the spin Hall conductivity (2.63 $\times$ 10$^5$ $\frac{\hbar}{2e}$ $Ω^{-1}$ m$^{-1}$) is found to be superior to values reported for other TMDs. The origin of this large damping-like SOT can be found in the interfacial properties of the TaS$_2$/Py heterostructure, and the experimental findings are complemented by the results from density functional theory calculations. The dominance of damping-like torque demonstrated in our study provides a promising path for designing next generation conducting TMD based low-powered quantum memory devices.

cond-mat.mes-hall↗

arXiv2020

SN 2010kd: Photometric and Spectroscopic Analysis of a Slow-Decaying Superluminous Supernova

This paper presents data and analysis of SN 2010kd, a low-redshift ($z = 0.101$) H-deficient superluminous supernova (SLSN), based on ultraviolet/optical photometry and optical spectroscopy spanning between $-$28 and +194 days relative to $\mathit{B}$ band maximum light. The $\mathit{B}$ band light curve comparison of SN 2010kd with a subset of well-studied SLSNe I at comparable redshifts indicates that it is a slow-decaying PTF12dam like SLSN. Analytical light-curve modeling using the $\mathtt{Minim}$ code suggests that the bolometric light curve of SN 2010kd favors circumstellar matter interaction for the powering mechanism. $\mathtt{SYNAPPS}$ modeling of the early-phase spectra does not identify broad H or He lines, whereas the photospheric-phase spectra are dominated by O I, O II, C II, C IV and Si II, particularly, presence of both low and high-velocity components of O II and Si II lines. The nebular-phase spectra of SN 2010kd are dominated by O I and Ca II emission lines similar to those seen in other SLSNe I. The line velocities in SN 2010kd exhibit flatter evolution curves similar to SN 2015bn but with comparatively higher values. SN 2010kd shows a higher single-zone local thermodynamic equilibrium temperature in comparison to PTF12dam and SN 2015bn, and it has an upper O I ejected mass limit of $\sim 10~M_\odot$. The host of SN 2010kd is a dwarf galaxy with a high star-formation rate ($\sim 0.18 \pm 0.04~M_\odot$ yr$^{-1}$) and extreme emission lines.

astro-ph.HE↗

arXiv2020

Fast Intent Classification for Spoken Language Understanding

Spoken Language Understanding (SLU) systems consist of several machine learning components operating together (e.g. intent classification, named entity recognition and resolution). Deep learning models have obtained state of the art results on several of these tasks, largely attributed to their better modeling capacity. However, an increase in modeling capacity comes with added costs of higher latency and energy usage, particularly when operating on low complexity devices. To address the latency and computational complexity issues, we explore a BranchyNet scheme on an intent classification scheme within SLU systems. The BranchyNet scheme when applied to a high complexity model, adds exit points at various stages in the model allowing early decision making for a set of queries to the SLU model. We conduct experiments on the Facebook Semantic Parsing dataset with two candidate model architectures for intent classification. Our experiments show that the BranchyNet scheme provides gains in terms of computational complexity without compromising model accuracy. We also conduct analytical studies regarding the improvements in the computational cost, distribution of utterances that egress from various exit points and the impact of adding more complexity to models with the BranchyNet scheme.

cs.CL↗

arXiv2019

Impact of the crystal orientation on spin-orbit torques in Fe/Pd bilayers

Spin-orbit torques in ferromagnetic (FM)/non-magnetic (NM) heterostructures offer more energy-efficient means to realize spin-logic devices; however, their strengths are determined by the heterostructure interface. This work examines crystal orientation impact on the spin-orbit torque efficiency in different Fe/Pd bilayer systems. Spin torque ferromagnetic measurements evidence that the damping-like torque efficiency is higher in epitaxial than in polycrystalline bilayer structures while the field-like torque is negligible in all bilayer structures. The strength of the damping-like torque decreases with deterioration of the bilayer epitaxial quality. The present finding provides fresh insight for the enhancement of spin-orbit torques in magnetic heterostructures.

cond-mat.mtrl-sci↗

arXiv2019

Modeling Feature Representations for Affective Speech using Generative Adversarial Networks

Emotion recognition is a classic field of research with a typical setup extracting features and feeding them through a classifier for prediction. On the other hand, generative models jointly capture the distributional relationship between emotions and the feature profiles. Relatively recently, Generative Adversarial Networks (GANs) have surfaced as a new class of generative models and have shown considerable success in modeling distributions in the fields of computer vision and natural language understanding. In this work, we experiment with variants of GAN architectures to generate feature vectors corresponding to an emotion in two ways: (i) A generator is trained with samples from a mixture prior. Each mixture component corresponds to an emotional class and can be sampled to generate features from the corresponding emotion. (ii) A one-hot vector corresponding to an emotion can be explicitly used to generate the features. We perform analysis on such models and also propose different metrics used to measure the performance of the GAN models in their ability to generate realistic synthetic samples. Apart from evaluation on a given dataset of interest, we perform a cross-corpus study where we study the utility of the synthetic samples as additional training data in low resource conditions.

cs.LG↗

arXiv2019

Extraordinary efficient spin-orbit torque switching in (W, Ta)/epitaxial-Co60Fe40/TiN heterostructures

The giant spin Hall effect in magnetic heterostructures along with low spin memory loss and high interfacial spin mixing conductance are prerequisites to realize energy efficient spin torque based logic devices. We report giant spin Hall angle (SHA) of 28.67 (5.09) for W (Ta) interfaced epi- Co60Fe40/TiN structures. The spin-orbit torque switching current density (J_Crit) is as low as 1.82 (8.21) MA/cm2 in W (Ta)/Co60Fe40(t_CoFe)/TiN structures whose origin lies in the epitaxial interfaces. These structures also exhibit very low spin memory loss and high spin mixing conductance. These extraordinary values of SHA and therefore ultra-low J_Crit in semiconducting industry compatible epitaxial materials combinations open up a new direction for the realization of energy efficient spin logic devices by utilizing epitaxial interfaces.

cond-mat.mtrl-sci↗

arXiv2019

Spin pumping and spin torque in interfacial tailored Co2FeAl/\b{eta}-Ta layers

The Heusler ferromagnetic (FM) compound Co2FeAl interfaced with a high-spin orbit coupling non-magnetic (NM) layer is a promising candidate for energy efficient spin logic circuits. The circuit potential depends on the strength of angular momentum transfer across the FM/NM interface; hence, requiring low spin memory loss and high spin-mixing conductance. To highlight this issue, spin pumping and spin-transfer torque ferromagnetic resonance measurements have been performed on Co_2FeAl/β-Ta heterostructures tailored with Cu interfacial layers. The interface tailored structure yields an enhancement of the effective spin-mixing conductance. The interface transparency and spin memory loss corrected values of the spin-mixing conductance, spin Hall angle and spin diffusion length are found to be 3.40 \pm 0.01 \times 10^{19} m^{-2}, 0.029 \pm 0.003, and 2.3 \pm 0.5 nm, respectively. Furthermore, a high current modulation of the effective damping of around 2.1 % has been achieved at an applied current density of 1 \times 10^9 A/m^2 , which clearly indicates the potential of using this heterostructure for energy efficient control in spin devices

cond-mat.mtrl-sci↗

arXiv2019

Engineering of spin mixing conductance in Ru/FeCo/Ru interfaces: Effect of Re Doping

We have deposited polycrystalline Re doped $(Fe_{65}Co_{35})_{100-x}Re_{x}$ (0 $\leq$ x $\leq$ 12.6 at\%) thin films grown under identical conditions and sandwiched between thin layers of Ru in order to study the phenomenon of spin pumping as a function of Re concentration. In-plane and out-of-plane ferromagnetic resonance spectroscopy results show an enhancement of the Gilbert damping with an increase in Re doping. We found evidence of an increase in the real part of effective spin mixing conductance [Re($g^{\uparrow\downarrow}_{eff}$)] with the increase in Re doping of 6.6 at\%, while a decrease is evident at higher Re doping. The increase in Re($g^{\uparrow\downarrow}_{eff}$) can be linked to the Re doping induced change of the interface electronic structure in the non-magnetic Ru layer and the effect interfacial spin-orbit coupling has on the effective spin-mixing conductance. The lowest and highest values of Re($g^{\uparrow\downarrow}_{eff}$) are found to be 9.883(02) $nm^{-2}$ and 19.697(02) $nm^{-2}$ for 0 at\% and 6.6 at\% Re doping, respectively. The saturation magnetization decreases with increasing Re doping, from 2.362(13) T for the undoped film to 1.740(03) T for 12.6 at\% Re doping. This study opens a new direction of tuning the spin-mixing conductance in magnetic heterostructures by doping of the ferromagnetic layerr, which is essential for the realization of energy efficient operation of spintronic devices.

cond-mat.mtrl-sci↗

arXiv2019

One-vs-All Models for Asynchronous Training: An Empirical Analysis

Any given classification problem can be modeled using multi-class or One-vs-All (OVA) architecture. An OVA system consists of as many OVA models as the number of classes, providing the advantage of asynchrony, where each OVA model can be re-trained independent of other models. This is particularly advantageous in settings where scalable model training is a consideration (for instance in an industrial environment where multiple and frequent updates need to be made to the classification system). In this paper, we conduct empirical analysis on realizing independent updates to OVA models and its impact on the accuracy of the overall OVA system. Given that asynchronous updates lead to differences in training datasets for OVA models, we first define a metric to quantify the differences in datasets. Thereafter, using Natural Language Understanding as a task of interest, we estimate the impact of three factors: (i) number of classes, (ii) number of data points and, (iii) divergences in training datasets across OVA models; on the OVA system accuracy. Finally, we observe the accuracy impact of increased asynchrony in a Spoken Language Understanding system. We analyze the results and establish that the proposed metric correlates strongly with the model performances in both the experimental settings.

cs.LG↗

arXiv2019

K-theory and 0-cycles on schemes

We prove Bloch's formula for 0-cycles on affine schemes over algebraically closed fields. We prove this formula also for projective schemes over algebraically closed fields which are regular in codimension one. Several applications, including Bloch's formula for 0-cycles with modulus, are derived.

math.AG↗

arXiv2019

Deep Learning for Bug-Localization in Student Programs

Providing feedback is an integral part of teaching. Most open online courses on programming make use of automated grading systems to support programming assignments and give real-time feedback. These systems usually rely on test results to quantify the programs' functional correctness. They return failing tests to the students as feedback. However, students may find it difficult to debug their programs if they receive no hints about where the bug is and how to fix it. In this work, we present the first deep learning based technique that can localize bugs in a faulty program w.r.t. a failing test, without even running the program. At the heart of our technique is a novel tree convolutional neural network which is trained to predict whether a program passes or fails a given test. To localize the bugs, we analyze the trained network using a state-of-the-art neural prediction attribution technique and see which lines of the programs make it predict the test outcomes. Our experiments show that the proposed technique is generally more accurate than two state-of-the-art program-spectrum based and one syntactic difference based bug-localization baselines.

cs.SE↗

arXiv2019

On evaluating CNN representations for low resource medical image classification

Convolutional Neural Networks (CNNs) have revolutionized performances in several machine learning tasks such as image classification, object tracking, and keyword spotting. However, given that they contain a large number of parameters, their direct applicability into low resource tasks is not straightforward. In this work, we experiment with an application of CNN models to gastrointestinal landmark classification with only a few thousands of training samples through transfer learning. As in a standard transfer learning approach, we train CNNs on a large external corpus, followed by representation extraction for the medical images. Finally, a classifier is trained on these CNN representations. However, given that several variants of CNNs exist, the choice of CNN is not obvious. To address this, we develop a novel metric that can be used to predict test performances, given CNN representations on the training set. Not only we demonstrate the superiority of the CNN based transfer learning approach against an assembly of knowledge driven features, but the proposed metric also carries an 87% correlation with the test set performances as obtained using various CNN representations.

eess.IV↗

arXiv2019

A multi-wavelength analysis of a collection of short-duration GRBs observed between 2012-2015

We investigate the prompt emission and the afterglow properties of short duration gamma-ray burst (sGRB) 130603B and another eight sGRB events during 2012-2015, observed by several multi-wavelength facilities including the GTC 10.4m telescope. Prompt emission high energy data of the events were obtained by INTEGRAL/SPI/ACS, Swift/BAT and Fermi/GBM satellites. The prompt emission data by INTEGRAL in the energy range of 0.1-10 MeV for sGRB 130603B, sGRB 140606A, sGRB 140930B, sGRB 141212A and sGRB 151228A do not show any signature of the extended emission or precursor activity and their spectral and temporal properties are similar to those seen in case of other short bursts. For sGRB130603B, our new afterglow photometric data constraints the pre jet-break temporal decay due to denser temporal coverage. For sGRB 130603B, the afterglow light curve, containing both our new as well as previously published photometric data is broadly consistent with the ISM afterglow model. Modeling of the host galaxies of sGRB 130603B and sGRB 141212A using the LePHARE software supports a scenario in which the environment of the burst is undergoing moderate star formation activity. From the inclusion of our late-time data for 8 other sGRBs we are able to; place tight constraints on the non-detection of the afterglow, host galaxy or any underlying kilonova emission. Our late-time afterglow observations of the sGRB 170817A/GW170817 are also discussed and compared with the sub-set of sGRBs.

astro-ph.HE↗

arXiv2019

Data augmentation for low resource sentiment analysis using generative adversarial networks

Sentiment analysis is a task that may suffer from a lack of data in certain cases, as the datasets are often generated and annotated by humans. In cases where data is inadequate for training discriminative models, generate models may aid training via data augmentation. Generative Adversarial Networks (GANs) are one such model that has advanced the state of the art in several tasks, including as image and text generation. In this paper, I train GAN models on low resource datasets, then use them for the purpose of data augmentation towards improving sentiment classifier generalization. Given the constraints of limited data, I explore various techniques to train the GAN models. I also present an analysis of the quality of generated GAN data as more training data for the GAN is made available. In this analysis, the generated data is evaluated as a test set (against a model trained on real data points) as well as a training set to train classification models. Finally, I also conduct a visual analysis by projecting the generated and the real data into a two-dimensional space using the t-Distributed Stochastic Neighbor Embedding (t-SNE) method.

cs.LG↗

arXiv2019

Direct Observation of Unusual Interfacial Dzyaloshinskii-Moriya Interaction in Graphene/NiFe/Ta Heterostructure

Graphene/ferromagnet interface promises a plethora of new science and technology. The interfacial Dzyaloshinskii Moriya interaction (iDMI) is essential for stabilizing chiral spin textures, which are important for future spintronic devices. Here, we report direct observation of iDMI in graphene/Ni80Fe20/Ta heterostructure from non-reciprocity in spin-wave dispersion using Brillouin light scattering (BLS) technique. Linear scaling of iDMI with the inverse of Ni80Fe20 thicknesses suggests primarily interfacial origin of iDMI. Both iDMI and spin-mixing conductance increase with the increase in defect density of graphene obtained by varying argon pressure during sputter deposition of Ni80Fe20. This suggests that the observed iDMI originates from defect-induced extrinsic spin-orbit coupling at the interface. The direct observation of iDMI at graphene/ferromagnet interface without perpendicular magnetic anisotropy opens new route in designing thin film heterostructures based on 2-D materials for controlling chiral spin structure such as skyrmions and bubbles, and magnetic domain-wall-based storage and memory devices.

cond-mat.mtrl-sci↗

arXiv2019

Verity: Blockchains to Detect Insider Attacks in DBMS

Integrity and security of the data in database systems are typically maintained with access control policies and firewalls. However, insider attacks -- where someone with an intimate knowledge of the system and administrative privileges tampers with the data -- pose a unique challenge. Measures like append only logging prove to be insufficient because an attacker with administrative privileges can alter logs and login records to eliminate the trace of attack, thus making insider attacks hard to detect. In this paper, we propose Verity -- first of a kind system to the best of our knowledge. Verity serves as a dataless framework by which any blockchain network can be used to store fixed-length metadata about tuples from any SQL database, without complete migration of the database. Verity uses a formalism for parsing SQL queries and query results to check the respective tuples' integrity using blockchains to detect insider attacks. We have implemented our technique using Hyperledger Fabric, Composer REST API, and SQLite database. Using TPC-H data and SQL queries of varying complexity and types, our experiments demonstrate that any overhead of integrity checking remains constant per tuple in a query's results, and scales linearly.

cs.DB↗

arXiv2018

A task in a suit and a tie: paraphrase generation with semantic augmentation

Paraphrasing is rooted in semantics. We show the effectiveness of transformers (Vaswani et al. 2017) for paraphrase generation and further improvements by incorporating PropBank labels via a multi-encoder. Evaluating on MSCOCO and WikiAnswers, we find that transformers are fast and effective, and that semantic augmentation for both transformers and LSTMs leads to sizable 2-3 point gains in BLEU, METEOR and TER. More importantly, we find surprisingly large gains on human evaluations compared to previous models. Nevertheless, manual inspection of generated paraphrases reveals ample room for improvement: even our best model produces human-acceptable paraphrases for only 28% of captions from the CHIA dataset (Sharma et al. 2018), and it fails spectacularly on sentences from Wikipedia. Overall, these results point to the potential for incorporating semantics in the task while highlighting the need for stronger evaluation.

cs.CL↗