arXiv subjects

Abhishek Kumar

Publications and source records attributed to Abhishek Kumar.

At least 181 records · Page 10Linked to original sources

arXiv2019

A Perspective on Recent Advances in Phosphorene Functionalization and its Application in Devices

Phosphorene, the 2D material derived from black phosphorus, has recently attracted a lot of interest for its properties, suitable for applications in material science. In particular, the physical features and the prominent chemical reactivity on its surface render this nanolayered substrate particularly promising for electrical and optoelectronic applications. In addition, being a new potential ligand for metals, it opens the way for a new role of the inorganic chemistry in the 2D world, with special reference to the field of catalysis. The aim of this review is to summarize the state of the art in this subject and to present our most recent results in preparation, functionalization and use of phosphorene and its decorated derivatives. In particular, we discuss several key points, which are currently under investigation: the synthesis, the characterization by theoretical calculations, the high pressure behaviour of black phosphorus, as well as decoration with nanoparticles and encapsulation in polymers. Finally, device fabrication and electrical transport measurements are overviewed on the basis of recent literature and new results collected in our laboratories.

cond-mat.mes-hall↗

arXiv2019

Energy Spectra and Fluxes in Dissipation Range of Turbulent and Laminar Flows

Two well-known turbulence models that describe the energy spectrum in the inertial and dissipative ranges simultaneously are by Pao~(1965) and Pope~(2000). In this paper, we compute the energy spectrum $E(k)$ and energy flux $Π(k)$ using direct numerical simulations on grids up to $4096^3$, and show consistency between the numerical results and the predictions by the aforementioned models. We also construct a model for laminar flows that predicts $E(k)\sim k^{-1} \exp(-k)$ and $Π(k)\sim k \exp(-k)$. Our model predictions match with the numerical results. We emphasize differences on the energy transfers in the two flows---they are {\em local} in the turbulent flows, and {\em nonlocal} in laminar flows.

physics.flu-dyn↗

arXiv2019

Stochastic bistable systems, and competing hysteresis and phase coexistence

In this paper we describe the solution of a stochastic bistable system from a dynamical perspective. We show how a single framework with variable noise can explain hysteresis at zero temperature and two-state coexistence in the presence of noise. This feature is similar to the phase transition of thermodynamics. Our mathematical model for bistable systems also explains how the width of a hysteresis loop shrinks in the presence of noise, and how variation in initial conditions can take such systems to different final states.

cond-mat.stat-mech↗

arXiv2019

Self- texturizing electronic-properties in a 2-dimensional GdAu2 layer on Au(111): the role of out-of-plane atomic displacement

Here, we show that the electronic properties of a surface-supported 2-dimensional (2D) layer structure can be self-texturized at the nanoscale. The local electronic properties are determined by structural relaxation processes through variable adsorption stacking configurations. We demonstrate that the spatially modulated layer-buckling, which arises from the lattice mismatch and the layer/substrate coupling at the GdAu2/Au(111) interface, is sufficient to locally open an energy gap of ~0.5eV at the Fermi level in an otherwise metallic layer. Additionally, this out-of-plane displacement of the Gd atoms patterns the character of the hybridized Gd-d states and shifts the center of mass of the Gd 4f multiplet proportionally to the lattice distortion. These findings demonstrate the close correlation between the electronic properties of the 2D-layer and its planarity. We demonstrate that the resulting template shows different chemical reactivity which may find important applications.

cond-mat.str-el↗

arXiv2019

A Scale Invariant Flatness Measure for Deep Network Minima

It has been empirically observed that the flatness of minima obtained from training deep networks seems to correlate with better generalization. However, for deep networks with positively homogeneous activations, most measures of sharpness/flatness are not invariant to rescaling of the network parameters, corresponding to the same function. This means that the measure of flatness/sharpness can be made as small or as large as possible through rescaling, rendering the quantitative measures meaningless. In this paper we show that for deep networks with positively homogenous activations, these rescalings constitute equivalence relations, and that these equivalence relations induce a quotient manifold structure in the parameter space. Using this manifold structure and an appropriate metric, we propose a Hessian-based measure for flatness that is invariant to rescaling. We use this new measure to confirm the proposition that Large-Batch SGD minima are indeed sharper than Small-Batch SGD minima.

stat.ML↗

arXiv2019

X-ray Diffraction Analysis of Cu2+ Doped Zn1-xCuxFe2O4 Spinel Nanoparticles using Williamson-Hall Plot Method

The nanoparticles (NPs) of Zn1-xCuxFe2O4 (ZCFO) spinels with x = 0, 0.2, 0.4, 0.6 and 0.8 were synthesized by a sol-gel combustion method using acetate precursor. The NPs of ZCFO were prepared by following calcination process at 600C for 8hrs. The synthesized NPs of ZCFO were characterized by X-ray diffraction (XRD) analysis using Rietveld refinement. The Rietveld refinement of the XRD patterns revealed that the ZCFO spinels crystallize into single diamond cubic structure with Fd-3m space group. The lattice constant and unit cell volume for ZCFO NPs shrink with enhancing doping concentration of Cu2+ ion. The crystalline growth in the NPs of ZCFO was examined by peak broadening present in the XRD pattern. The Williamson-Hall (W-H) plot method were used to study the individual involvements of crystallite sizes and lattice strain on the peak broadening of the NPs of ZCFO spinels. Whereas, particle size of the ZCFO sample with x = 0.40 was estimated by high-resolution scanning electron microscopy micrographs

cond-mat.mtrl-sci↗

arXiv2019

BlockDrop: Dynamic Inference Paths in Residual Networks

Very deep convolutional neural networks offer excellent recognition results, yet their computational expense limits their impact for many real-world applications. We introduce BlockDrop, an approach that learns to dynamically choose which layers of a deep network to execute during inference so as to best reduce total computation without degrading prediction accuracy. Exploiting the robustness of Residual Networks (ResNets) to layer dropping, our framework selects on-the-fly which residual blocks to evaluate for a given novel image. In particular, given a pretrained ResNet, we train a policy network in an associative reinforcement learning setting for the dual reward of utilizing a minimal number of blocks while preserving recognition accuracy. We conduct extensive experiments on CIFAR and ImageNet. The results provide strong quantitative and qualitative evidence that these learned policies not only accelerate inference but also encode meaningful visual information. Built upon a ResNet-101 model, our method achieves a speedup of 20\% on average, going as high as 36\% for some images, while maintaining the same 76.4\% top-1 accuracy on ImageNet.

cs.CV↗

arXiv2018

Variational Inference of Disentangled Latent Concepts from Unlabeled Observations

Disentangled representations, where the higher level data generative factors are reflected in disjoint latent dimensions, offer several benefits such as ease of deriving invariant representations, transferability to other tasks, interpretability, etc. We consider the problem of unsupervised learning of disentangled representations from large pool of unlabeled observations, and propose a variational inference based approach to infer disentangled latent factors. We introduce a regularizer on the expectation of the approximate posterior over observed data that encourages the disentanglement. We also propose a new disentanglement metric which is better aligned with the qualitative disentanglement observed in the decoder's output. We empirically observe significant improvement over existing methods in terms of both disentanglement and data likelihood (reconstruction quality).

cs.LG↗

arXiv2018

Polaron mobility in the "beyond quasiparticles" regime

In a number of physical situations, from polarons to Dirac liquids and to non-Fermi liquids, one encounters the "beyond quasiparticles" regime, in which the inelastic scattering rate exceeds the thermal energy of quasiparticles. Transport in this regime cannot be described by the kinetic equation. We employ the Diagrammatic Monte Carlo method to study the mobility of a Fröhlich polaron in this regime and discover a number of non-perturbative effects: a strong violation of the Mott-Ioffe-Regel criterion at intermediate and strong couplings, a mobility minimum at $T \sim Ω$ in the strong-coupling limit ($Ω$ is the optical mode frequency), a substantial delay in the onset of an exponential dependence of the mobility for $T<Ω$ at intermediate coupling, and complete smearing of the Drude peak at strong coupling. These effects should be taken into account when interpreting mobility data in materials with strong electron-phonon coupling.

cond-mat.str-el↗

arXiv2018

Understanding Unequal Gender Classification Accuracy from Face Images

Recent work shows unequal performance of commercial face classification services in the gender classification task across intersectional groups defined by skin type and gender. Accuracy on dark-skinned females is significantly worse than on any other group. In this paper, we conduct several analyses to try to uncover the reason for this gap. The main finding, perhaps surprisingly, is that skin type is not the driver. This conclusion is reached via stability experiments that vary an image's skin type via color-theoretic methods, namely luminance mode-shift and optimal transport. A second suspect, hair length, is also shown not to be the driver via experiments on face images cropped to exclude the hair. Finally, using contrastive post-hoc explanation techniques for neural networks, we bring forth evidence suggesting that differences in lip, eye and cheek structure across ethnicity lead to the differences. Further, lip and eye makeup are seen as strong predictors for a female face, which is a troubling propagation of a gender stereotype.

cs.CV↗

arXiv2018

Delta-encoder: an effective sample synthesis method for few-shot object recognition

Learning to classify new categories based on just one or a few examples is a long-standing challenge in modern computer vision. In this work, we proposes a simple yet effective method for few-shot (and one-shot) object recognition. Our approach is based on a modified auto-encoder, denoted Delta-encoder, that learns to synthesize new samples for an unseen category just by seeing few examples from it. The synthesized samples are then used to train a classifier. The proposed approach learns to both extract transferable intra-class deformations, or "deltas", between same-class pairs of training examples, and to apply those deltas to the few provided examples of a novel class (unseen during training) in order to efficiently synthesize samples from that new class. The proposed method improves over the state-of-the-art in one-shot object-recognition and compares favorably in the few-shot case. Upon acceptance code will be made available.

cs.CV↗

arXiv2018

SpotTune: Transfer Learning through Adaptive Fine-tuning

Transfer learning, which allows a source task to affect the inductive bias of the target task, is widely used in computer vision. The typical way of conducting transfer learning with deep neural networks is to fine-tune a model pre-trained on the source task using data from the target task. In this paper, we propose an adaptive fine-tuning approach, called SpotTune, which finds the optimal fine-tuning strategy per instance for the target data. In SpotTune, given an image from the target task, a policy network is used to make routing decisions on whether to pass the image through the fine-tuned layers or the pre-trained layers. We conduct extensive experiments to demonstrate the effectiveness of the proposed approach. Our method outperforms the traditional fine-tuning approach on 12 out of 14 standard datasets.We also compare SpotTune with other state-of-the-art fine-tuning strategies, showing superior performance. On the Visual Decathlon datasets, our method achieves the highest score across the board without bells and whistles.

cs.CV↗

arXiv2018

MT-CGCNN: Integrating Crystal Graph Convolutional Neural Network with Multitask Learning for Material Property Prediction

Developing accurate, transferable and computationally inexpensive machine learning models can rapidly accelerate the discovery and development of new materials. Some of the major challenges involved in developing such models are, (i) limited availability of materials data as compared to other fields, (ii) lack of universal descriptor of materials to predict its various properties. The limited availability of materials data can be addressed through transfer learning, while the generic representation was recently addressed by Xie and Grossman [1], where they developed a crystal graph convolutional neural network (CGCNN) that provides a unified representation of crystals. In this work, we develop a new model (MT-CGCNN) by integrating CGCNN with transfer learning based on multi-task (MT) learning. We demonstrate the effectiveness of MT-CGCNN by simultaneous prediction of various material properties such as Formation Energy, Band Gap and Fermi Energy for a wide range of inorganic crystals (46774 materials). MT-CGCNN is able to reduce the test error when employed on correlated properties by upto 8%. The model prediction has lower test error compared to CGCNN, even when the training data is reduced by 10%. We also demonstrate our model's better performance through prediction of end user scenario related to metal/non-metal classification. These results encourage further development of machine learning approaches which leverage multi-task learning to address the aforementioned challenges in the discovery of new materials. We make MT-CGCNN's source code available to encourage reproducible research.

cs.LG↗

arXiv2018

Co-regularized Alignment for Unsupervised Domain Adaptation

Deep neural networks, trained with large amount of labeled data, can fail to generalize well when tested with examples from a \emph{target domain} whose distribution differs from the training data distribution, referred as the \emph{source domain}. It can be expensive or even infeasible to obtain required amount of labeled data in all possible domains. Unsupervised domain adaptation sets out to address this problem, aiming to learn a good predictive model for the target domain using labeled examples from the source domain but only unlabeled examples from the target domain. Domain alignment approaches this problem by matching the source and target feature distributions, and has been used as a key component in many state-of-the-art domain adaptation methods. However, matching the marginal feature distributions does not guarantee that the corresponding class conditional distributions will be aligned across the two domains. We propose co-regularized domain alignment for unsupervised domain adaptation, which constructs multiple diverse feature spaces and aligns source and target distributions in each of them individually, while encouraging that alignments agree with each other with regard to the class predictions on the unlabeled target examples. The proposed method is generic and can be used to improve any domain adaptation method which uses domain alignment. We instantiate it in the context of a recent state-of-the-art method and observe that it provides significant performance improvements on several domain adaptation benchmarks.

cs.LG↗

arXiv2018

Phase change-photonic framework for terahertz wave control

The advancement in capabilities enabled by handheld devices and the prospective internet of things has led to an explosion of wireless data consumption in the last decade. The ever-increasing demand for high speed data transfer has projected the unprecedented importance of terahertz spectral range for next generation wireless communication systems. However, versatile interaction and active manipulation of terahertz waves required for a wide-range of terahertz technologies are either unavailable or unscalable. Here, we report on an integrated platform of chalcogenide phase change material, germanium antimony telluride (GST), with metamaterials for multidimensional manipulation of terahertz waves. The thermal, electrical and optical stimulation of GST provide multilevel non-volatility, sub-wavelength spatial selectivity and variable ultrafast switching of terahertz waves, respectively. These multifaceted response of phase change material can be used in tandem with metamaterial that enables strong field confinement along with on-demand spectral response and frequency selectivity. The unique and multifarious properties of GST combined with the versatility and scalability of metamaterial design and ease of fabrication could aid in the development of advanced devices operating in terahertz range, such as spatio-temporal light modulators, neuromorphic photonic devices, compressive imagers, and switchable holograms, vital for the next generation wireless communication systems.

physics.app-ph↗

arXiv2018

Required sensitivity to search the neutrinoless double beta decay in $^{124}Sn$

\textbf{T}he \textbf{IN}dia's \textbf{TIN} (TIN.TIN) detector is under development in the search for neutrinoless double-$β$ decay (0$νββ$) using 90\% enriched $^{124}$Sn isotope as the target mass. This detector will be housed in the upcoming underground facility of the \textbf{I}ndia based \textbf{N}eutrino \textbf{O}bservatory. We present the most important experimental parameters that would be used in the study of required sensitivity for the TIN.TIN experiment to probe the neutrino mass hierarchy. The sensitivity of the TIN.TIN detector in the presence of sole two neutrino double-$β$ decay (2$νββ$) decay background is studied at various energy resolutions. The most optimistic and pessimistic scenario to probe the neutrino mass hierarchy at 3$σ$ sensitivity level and 90\% C.L. is also discussed.

hep-ph↗

arXiv2018

STM Study of Exfoliated Few Layer Black Phosphorus Annealed in Ultrahigh Vacuum

Black Phosphorus (bP) has emerged as an interesting addition to the category of two-dimensional materials. Surface-science studies on this material are of great interest, but they are hampered by bP's high reactivity to oxygen and water, a major challenge to scanning tunneling microscopy (STM) experiments. As a consequence, the large majority of these studies were performed by cleaving a bulk crystal in situ. Here we present a study of surface modifications on exfoliated bP flakes upon consecutive annealing steps, up to 550 C, well above the sublimation temperature of bP. In particular, our attention is focused on the temperature range 375 C - 400 C, when sublimation starts, and a controlled desorption from the surface occurs alongside with the formation of characteristic well-aligned craters. There is an open debate in the literature about the crystallographic orientation of these craters, whether they align along the zigzag or the armchair direction. Thanks to the atomic resolution provided by STM, we are able to identify the orientation of the craters with respect to the bP crystal: the long axis of the craters is aligned along the zigzag direction of bP. This allows us to solve the controversy, and, moreover, to provide insight in the underlying desorption mechanism leading to crater formation.

cond-mat.mes-hall↗

arXiv2018

Knowledge-enriched Two-layered Attention Network for Sentiment Analysis

We propose a novel two-layered attention network based on Bidirectional Long Short-Term Memory for sentiment analysis. The novel two-layered attention network takes advantage of the external knowledge bases to improve the sentiment prediction. It uses the Knowledge Graph Embedding generated using the WordNet. We build our model by combining the two-layered attention network with the supervised model based on Support Vector Regression using a Multilayer Perceptron network for sentiment analysis. We evaluate our model on the benchmark dataset of SemEval 2017 Task 5. Experimental results show that the proposed model surpasses the top system of SemEval 2017 Task 5. The model performs significantly better by improving the state-of-the-art system at SemEval 2017 Task 5 by 1.7 and 3.7 points for sub-tracks 1 and 2 respectively.

cs.CL↗