arXiv subjects

Gaurav Kumar

Publications and source records attributed to Gaurav Kumar.

62 records · Page 4Linked to original sources

arXiv2019

Reinforcement Learning based Curriculum Optimization for Neural Machine Translation

We consider the problem of making efficient use of heterogeneous training data in neural machine translation (NMT). Specifically, given a training dataset with a sentence-level feature such as noise, we seek an optimal curriculum, or order for presenting examples to the system during training. Our curriculum framework allows examples to appear an arbitrary number of times, and thus generalizes data weighting, filtering, and fine-tuning schemes. Rather than relying on prior knowledge to design a curriculum, we use reinforcement learning to learn one automatically, jointly with the NMT system, in the course of a single training run. We show that this approach can beat uniform and filtering baselines on Paracrawl and WMT English-to-French datasets by up to +3.4 BLEU, and match the performance of a hand-designed, state-of-the-art curriculum.

cs.CL↗

arXiv2018

An Empirical Exploration of Curriculum Learning for Neural Machine Translation

Machine translation systems based on deep neural networks are expensive to train. Curriculum learning aims to address this issue by choosing the order in which samples are presented during training to help train better models faster. We adopt a probabilistic view of curriculum learning, which lets us flexibly evaluate the impact of curricula design, and perform an extensive exploration on a German-English translation task. Results show that it is possible to improve convergence time at no loss in translation quality. However, results are highly sensitive to the choice of sample difficulty criteria, curriculum schedule and other hyperparameters.

cs.CL↗

arXiv2018

A Phonon Boltzmann Study of Microscale Thermal Transport in $α$-RDX Cook-Off

The microscale thermal transport properties of $α$RDX are believed to be major factors in the initiation process. In this study we present a thorough examination of phonon properties which dominate energy storage and transport in $α$RDX. The phonon lifetimes are determined for all phonon branches, revealing the characteristic time scale of energy transfer amongst phonon modes. The phonon parameters also serve as inputs to a full Brillouin zone three dimensional phonon transport simulation in the presence of a hotspot. In addition to identifying the phonon mode contributions to thermal transport, and as N-N bond breaking is integral to disassociation, we identify phonon modes corresponding to large N-N bond stretch analyzing the manner in which these modes store and transfer energy.

cond-mat.mtrl-sci↗

arXiv2018

Binary Multifunctional Ultrabroadband Self-Powered g-C3N4 /Si Heterojunction High-Speed Photodetector

Compact optical detectors with fast binary photoswitching over a broad range of wavelength are essential as an interconnect for any light-based parallel, real-time computing. Despite of the tremendous technological advancements yet there is no such single device available that meets the specifications. Here we report a multifunctional self-powered high-speed ultrabroadband (250-1650 nm) photodetector based on g-C3N4/Si hybrid 2D/3D structure. The device shows a novel binary photoswitching (change in current from positive to negative) in response to OFF/ON light illumination at small forward bias (<0.1 V) covering 250-1350 nm. At zero bias, the device displays an extremely high ON/OFF ratio of 1.2 x 10^5 under 680 nm (49 microWcm-2) illumination. The device also shows an ultrasensitive behaviour over the entire operating range at low light illuminations, with highest responsivity (1.2 AW-1), detectivity (2.8 x 10^14 Jones) and external quantum efficiency (213%) at 680 nm. The response and recovery speeds are typically 0.23 and 0.60 ms, respectively, under 288 Hz light switching frequency. Dramatically improved performance of our device is attributed to the heterojunctions formed by the ultrathin g-C3N4 nanosheets embedded in the Si surface.

physics.app-ph↗

arXiv2018

Tubulation pattern of membrane vesicles coated with bio filaments

Narrow membrane tubes are commonly pulled out from the surface of phospholipid vesicles using forces applied either through laser or magnetic tweezers or through the action of processive motor proteins. Recent examples have emerged where such tubes spontaneously grow from vesicles coated with bioactive cytoskeletal filaments (e.g. FtsZ, microtubule) in the presence GTP. We show how a soft vesicle deforms due to the interplay between its topology, local curvature and the forces due to the active filaments. We present results from Dynamically Triangulated Monte Carlo simulations of a spherical continuum membrane coated with a nematic field and show how the intrinsic curvature of the filaments and their ordering interactions drive membrane tubulation. We predict interesting patterns of nematic defects, on curved 2D membrane surfaces, which promote tube formation. Implication of our model for more dynamic cases where vesicles coated with an active mixture of microtubule and myosin show shape oscillation, are also discussed. All these cases point to a common theme that defect locations on 2D membrane surfaces are hot spots of membrane deformation activity.

physics.bio-ph↗

arXiv2017

Using of heterogeneous corpora for training of an ASR system

The paper summarizes the development of the LVCSR system built as a part of the Pashto speech-translation system at the SCALE (Summer Camp for Applied Language Exploration) 2015 workshop on "Speech-to-text-translation for low-resource languages". The Pashto language was chosen as a good "proxy" low-resource language, exhibiting multiple phenomena which make the speech-recognition and and speech-to-text-translation systems development hard. Even when the amount of data is seemingly sufficient, given the fact that the data originates from multiple sources, the preliminary experiments reveal that there is little to no benefit in merging (concatenating) the corpora and more elaborate ways of making use of all of the data must be worked out. This paper concentrates only on the LVCSR part and presents a range of different techniques that were found to be useful in order to benefit from multiple different corpora

cs.CL↗

arXiv2017

DyNet: The Dynamic Neural Network Toolkit

We describe DyNet, a toolkit for implementing neural network models based on dynamic declaration of network structure. In the static declaration strategy that is used in toolkits like Theano, CNTK, and TensorFlow, the user first defines a computation graph (a symbolic representation of the computation), and then examples are fed into an engine that executes this computation and computes its derivatives. In DyNet's dynamic declaration strategy, computation graph construction is mostly transparent, being implicitly constructed by executing procedural code that computes the network outputs, and the user is free to use different network structures for each input. Dynamic declaration thus facilitates the implementation of more complicated network architectures, and DyNet is specifically designed to allow users to implement their models in a way that is idiomatic in their preferred programming language (C++ or Python). One challenge with dynamic declaration is that because the symbolic computation graph is defined anew for every training example, its construction must have low overhead. To achieve this, DyNet has an optimized C++ backend and lightweight graph representation. Experiments show that DyNet's speeds are faster than or comparable with static declaration toolkits, and significantly faster than Chainer, another dynamic declaration toolkit. DyNet is released open-source under the Apache 2.0 license and available at http://github.com/clab/dynet.

stat.ML↗

arXiv2016

Ultrasensitive Self-powered large area planar GaN UV-photodetector using reduced graphene oxide electrodes

A simplistic design of a self-powered UV-photodetector device based on hybrid r-GO/GaN is demonstrated. Under zero bias, the fabricated hybrid photodetector shows a photosensivity of ~ 85% while ohmic contact GaN photodetector with identical device structure exhibits only ~ 5.3% photosensivity at 350 nm illumination (18 microWatt/cm^2). The responsivity and detectivity of the hybrid device were found to be 1.54 mA/W and 1.45x10^10 Jones (cm Hz^(1/2) W^(-1)), respectively at zero bias under 350 nm illumination (18 microWatt/cm^2) with fast response (60 ms), recovery time (267 ms) and excellent repeatability. Power density-dependent responsivity & detectivity revealed ultrasensitive behaviour under low light conditions. The source of observed self-powered effect in hybrid photodetector is attributed to the depletion region formed at the r-GO and GaN quasi-ohmic interface.

cond-mat.mtrl-sci↗