arXiv subjects

Xinyu Liu

Publications and source records attributed to Xinyu Liu.

248 records · Page 14Linked to original sources

arXiv2019

Learning Efficient Lexically-Constrained Neural Machine Translation with External Memory

Recent years has witnessed dramatic progress of neural machine translation (NMT), however, the method of manually guiding the translation procedure remains to be better explored. Previous works proposed to handle such problem through lexcially-constrained beam search in the decoding phase. Unfortunately, these lexically-constrained beam search methods suffer two fatal disadvantages: high computational complexity and hard beam search which generates unexpected translations. In this paper, we propose to learn the ability of lexically-constrained translation with external memory, which can overcome the above mentioned disadvantages. For the training process, automatically extracted phrase pairs are extracted from alignment and sentence parsing, then further be encoded into an external memory. This memory is then used to provide lexically-constrained information for training through a memory-attention machanism. Various experiments are conducted on WMT Chinese to English and English to German tasks. All the results can demonstrate the effectiveness of our method.

cs.CL↗

arXiv2018

Spelling Error Correction Using a Nested RNN Model and Pseudo Training Data

We propose a nested recurrent neural network (nested RNN) model for English spelling error correction and generate pseudo data based on phonetic similarity to train it. The model fuses orthographic information and context as a whole and is trained in an end-to-end fashion. This avoids feature engineering and does not rely on a noisy channel model as in traditional methods. Experiments show that the proposed method is superior to existing systems in correcting spelling errors.

cs.CL↗

arXiv2018

Dex-Net 3.0: Computing Robust Robot Vacuum Suction Grasp Targets in Point Clouds using a New Analytic Model and Deep Learning

Vacuum-based end effectors are widely used in industry and are often preferred over parallel-jaw and multifinger grippers due to their ability to lift objects with a single point of contact. Suction grasp planners often target planar surfaces on point clouds near the estimated centroid of an object. In this paper, we propose a compliant suction contact model that computes the quality of the seal between the suction cup and local target surface and a measure of the ability of the suction grasp to resist an external gravity wrench. To characterize grasps, we estimate robustness to perturbations in end-effector and object pose, material properties, and external wrenches. We analyze grasps across 1,500 3D object models to generate Dex-Net 3.0, a dataset of 2.8 million point clouds, suction grasps, and grasp robustness labels. We use Dex-Net 3.0 to train a Grasp Quality Convolutional Neural Network (GQ-CNN) to classify robust suction targets in point clouds containing a single object. We evaluate the resulting system in 350 physical trials on an ABB YuMi fitted with a pneumatic suction gripper. When evaluated on novel objects that we categorize as Basic (prismatic or cylindrical), Typical (more complex geometry), and Adversarial (with few available suction-grasp points) Dex-Net 3.0 achieves success rates of 98$\%$, 82$\%$, and 58$\%$ respectively, improving to 81$\%$ in the latter case when the training set includes only adversarial objects. Code, datasets, and supplemental material can be found at http://berkeleyautomation.github.io/dex-net .

cs.RO↗

arXiv2017

Dex-Net 2.0: Deep Learning to Plan Robust Grasps with Synthetic Point Clouds and Analytic Grasp Metrics

To reduce data collection time for deep learning of robust robotic grasp plans, we explore training from a synthetic dataset of 6.7 million point clouds, grasps, and analytic grasp metrics generated from thousands of 3D models from Dex-Net 1.0 in randomized poses on a table. We use the resulting dataset, Dex-Net 2.0, to train a Grasp Quality Convolutional Neural Network (GQ-CNN) model that rapidly predicts the probability of success of grasps from depth images, where grasps are specified as the planar position, angle, and depth of a gripper relative to an RGB-D sensor. Experiments with over 1,000 trials on an ABB YuMi comparing grasp planning methods on singulated objects suggest that a GQ-CNN trained with only synthetic data from Dex-Net 2.0 can be used to plan grasps in 0.8sec with a success rate of 93% on eight known objects with adversarial geometry and is 3x faster than registering point clouds to a precomputed dataset of objects and indexing grasps. The Dex-Net 2.0 grasp planner also has the highest success rate on a dataset of 10 novel rigid objects and achieves 99% precision (one false positive out of 69 grasps classified as robust) on a dataset of 40 novel household objects, some of which are articulated or deformable. Code, datasets, videos, and supplementary material are available at http://berkeleyautomation.github.io/dex-net .

cs.RO↗

arXiv2017

MBE growth of 2H-MoTe2 and 1T'-MoTe2 on 3D substrates

MoTe2 is the least explored material in the Molybdenum-chalcogen family, which crystallizes in thermodynamically stable semiconducting 2H phase at \textless 500 C and 1T' metallic phase at higher temperatures. Molecular beam epitaxy (MBE) provides an unique opportunity to tackle the small electronegativity difference between Mo and Te while growing layer by layer away from thermodynamic equilibrium. For a few-layer MoTe2 grown at a moderate rate of $\sim$6 mins per monolayer under varied Te:Mo flux ratio and substrate temperature, the boundary between the 2 phases in MBE grown MoTe2 on CaF2 is characterized using Reflection high-energy electron diffraction (RHEED), Raman spectroscopy and X-ray photoemission spectroscopy (XPS). Grazing incidence X-ray diffraction (GI-XRD) reveals a grain size of $\sim$90 Å and presence of twinned grains. XRD, transmission electron miscroscopy, RHEED, low energy electron diffraction along with lack of electrical conductivity modulation by field effect in MBE 2H-MoTe2 on GaAs (111) B show likelihood of excess Te incorporation in the films. Finally, thermal stability and air sensitivity of MBE 2H-MoTe2 is investigated by temperature dependent XRD and XPS, respectively.

cond-mat.mtrl-sci↗

arXiv2017

Observation of oscillatory relaxation in the Sn-terminated surface of epitaxial rock-salt SnSe $\{111\}$ topological crystalline insulator

Topological crystalline insulators have been recently predicted and observed in rock-salt structure SnSe $\{111\}$ thin films. Previous studies have suggested that the Se-terminated surface of this thin film with hydrogen passivation, has a reduced surface energy and is thus a preferred configuration. In this paper, synchrotron-based angle-resolved photoemission spectroscopy, along with density functional theory calculations, are used to demonstrate conclusively that a rock-salt SnSe $\{111\}$ thin film epitaxially-grown on \ce{Bi2Se3} has a stable Sn-terminated surface. These observations are supported by low energy electron diffraction (LEED) intensity-voltage measurements and dynamical LEED calculations, which further show that the Sn-terminated SnSe $\{111\}$ thin film has undergone a surface structural relaxation of the interlayer spacing between the Sn and Se atomic planes. In sharp contrast to the Se-terminated counterpart, the observed Dirac surface state in the Sn-terminated SnSe $\{111\}$ thin film is shown to yield a high Fermi velocity, $0.50\times10^6$m/s, which suggests a potential mechanism of engineering the Dirac surface state of topological materials by tuning the surface configuration.

cond-mat.mtrl-sci↗

arXiv2016

Four-wave mixing in perovskite photovoltaic materials reveals long dephasing times and weaker many-body interactions than GaAs

Perovksite semiconductors have shown promise for low-cost solar cells, lasers and photodetectors, yet their fundamental photophysical properties are not well understood. Recent observations of a low ($\sim$few meV) exciton binding energy and evidence of hot phonon effects in the room temperature phase suggest that perovskites are much closer to inorganic semiconductors than the absorber layers in traditional organic photovoltaics, signaling the need for experiments that shed light on the placement of perovskite materials within the spectrum of semiconductors used in optoelectronics and photovoltaics. Here we use four-wave mixing (FWM) to contrast the coherent optical response of CH$_3$NH$_3$PbI$_3$ thin films and crystalline GaAs. At carrier densities relevant for solar cell operation, our results show that carriers interact surprisingly weakly via the Coulomb interaction in perovskite, much weaker than in inorganic semiconductors. These weak many-body effects lead to a dephasing time in CH$_3$NH$_3$PbI$_3$ $\sim$3 times longer than in GaAs. Our results also show that the strong enhancement of the exciton FWM signal tied to excitation-induced dephasing in GaAs and other III-V semiconductors does not occur in perovskite due to weak exciton-carrier scattering interactions.

cond-mat.mtrl-sci↗

arXiv2015

Collective magnetization dynamics in ferromagnetic (Ga,Mn)As mediated by photo-excited carriers

We present a study of photo-excited magnetization dynamics in ferromagnetic (Ga,Mn)As films observed by time-resolved magneto-optical measurements. The magnetization precession triggered by linearly polarized optical pulses in the absence of an external field shows a strong dependence on photon frequency when the photo-excitation energy approaches the band-edge of (Ga,Mn)As. This can be understood in terms of magnetic anisotropy modulation by both laser heating of the sample and by hole-induced non-thermal paths. Our findings provide a means for identifying the transition of laser-triggered magnetization dynamics from thermal to non-thermal mechanisms, a result that is of importance for ultrafast optical spin manipulation in ferromagnetic materials via non-thermal paths.

cond-mat.mes-hall↗

arXiv2014

Molecular beam epitaxial growth of MoSe2 on graphite, CaF2 and graphene

We report the structural and optical properties of molecular beam epitaxy (MBE) grown 2-dimensional (2D) material molybdenum diselenide (MoSe2) on graphite, CaF2 and epitaxial graphene. Extensive characterizations reveal that 2H- MoSe2 grows by van-der-Waals epitaxy on all 3 substrates with a preferred crystallographic orientation and a Mo:Se ratio of 1:2. Photoluminescence at room temperature (~1.56 eV) is observed in monolayer MoSe2 on both CaF2 and epitaxial graphene. The band edge absorption is very sharp, <60 meV over 3 decades. Overcoming the observed small grains by promoting mobility of Mo atoms would make MBE a powerful technique to achieve high quality 2D materials and heterostructures.

cond-mat.mtrl-sci↗

arXiv2014

Ultrafast Probes of Nonequilibrium Hole Spin Relaxation in Ferromagnetic Semiconductor GaMnAs

We directly measure the hole spin lifetime in ferromagnetic GaMnAs via time- and polarization-resolved spectroscopy. Below the Curie temperature Tc, an ultrafast photoexcitation with linearly-polarized light is shown to create a non-equilibrium hole spin population via the dynamical polarization of holes through p-d exchange scattering with ferromagnetically-ordered Mn spins, and we characterize their relaxation dynamics. The observed relaxation consists of a distinct three-step recovery : (i) femtosecond (fs) hole spin relaxation ~ $160-200 fs, (ii) picosecond (ps) hole energy relaxation ~ 1-2 ps, and (iii) a coherent, damped Mn spin precession with a period of ~ 250 ps. The transient amplitude of the hole spin component diminishes with increasing temperature, directly following the ferromagnetic order, while the hole energy amplitude shows negligible temperature change, consistent with our interpretation. Our results thus establish the hole spin lifetimes in ferromagnetic semiconductors and demonstrate a novel spectroscopy method for studying non-equilibrium hole spins in the presence of correlation and magnetic order.

cond-mat.mtrl-sci↗

arXiv2012

Observation of the fractional ac Josephson effect: the signature of Majorana particles

Topological superconductors which support Majorana fermions are thought to be realized in one-dimensional semiconducting wires coupled to a superconductor. Such excitations are expected to exhibit non-Abelian statistics and can be used to realize quantum gates that are topologically protected from local sources of decoherence. Here we report the observation of the fractional a.c. Josephson effect in a hybrid semiconductor/superconductor InSb/Nb nanowire junction, a hallmark of topological matter. When the junction is irradiated with a radio-frequency f in the absence of an external magnetic field, quantized voltage steps (Shapiro steps) with a height hf/2e are observed, as is expected for conventional superconductor junctions, where the supercurrent is carried by charge-2e Cooper pairs. At high magnetic fields the height of the first Shapiro step is doubled to hf/e, suggesting that the supercurrent is carried by charge-e quasiparticles. This is a unique signature of Majorana fermions, elusive particles predicted ca. 80 years ago.

cond-mat.mes-hall↗

arXiv2011

dMotifGreedy: a novel tool for de novo discovery of DNA motifs with enhanced power of reporting distinct motifs

De novo discovery of over-represented DNA motifs is one of the major challenges in computational biology. Although numerous tools have been available for de novo motif discovery, many of these tools are subject to local optima phenomena, which may hinder detection of multiple distinct motifs. A greedy algorithm based tool named dMotifGreedy was developed. dMotifGreedy begins by searching for candidate motifs from pair-wise local alignments of input sequences and then computes an optimal global solution for each candidate motif through a greedy algorithm. dMotifGreedy has competitive performance in detecting a true motif and greatly enhanced performance in detecting multiple distinct true motifs. dMotifGreedy is freely available via a stand-alone program at http://lambchop.ads.uga.edu/dmotifgreedy/download.php.

q-bio.GN↗

arXiv2007

Magnetic Anisotropy, Spin Pinning and Exchange Constants of (Ga,Mn)As films

We present a detailed investigation of exchange-dominated nonpropagating spin-wave modes in a series of 100 nm Ga$_{1 - x}$Mn$_{x}$As films with Mn concentrations $x$ ranging from 0.02 to 0.08. The angular and Mn concentration dependences of spin wave resonance modes have been studied for both as-grown and annealed samples. Our results indicate that the magnetic anisotropy terms of Ga$_{1 - x}$Mn$_{x}$As depend on the Mn concentration $x$, but are also strongly affected by sample growth conditions; moreover, the magnetic anisotropy of Ga$_{1 - x}$Mn$_{x}$As films is found to be clearly linked to the Curie temperature. The spin wave resonance spectra consist of a series of well resolved standing spin-wave modes. The observed mode patterns are consistent with the Portis volume-inhomogeneity model, in which a spatially nonuniform anisotropy field acts on the Mn spins. The analysis of these exchange-dominated spin wave modes, including their angular dependences, allows us to establish the exchange stiffness constants for Ga$_{1 - x}$Mn$_{x}$As films.

cond-mat.mtrl-sci↗

arXiv2002

Predictive Ability of QCD Sum Rules for Excited Baryons

The masses of octet baryons are calculated by the method of QCD sum rules. Using generalized interpolating fields, three independent sets of QCD sum rules are derived which allow the extraction of low-lying N* states with spin-parity 1/2+, 1/2- and 3/2- in both the non-strange and strange channels. The predictive ability of the sum rules is examined by a Monte-Carlo based analysis procedure in which the three phenomenological parameters (mass, coupling, threshold) are treated as free parameters simultaneously. Realistic uncertainties in these parameters are obtained by simultaneously exploring all uncertainties in the QCD input parameters. Those sum rules with good predictive power are identified and their predictions are compared with experiment where available.

nucl-th↗