arXiv subjects

Cheng Wang

Publications and source records attributed to Cheng Wang.

At least 433 records · Page 24Linked to original sources

arXiv2019

State-Regularized Recurrent Neural Networks

Recurrent neural networks are a widely used class of neural architectures. They have, however, two shortcomings. First, it is difficult to understand what exactly they learn. Second, they tend to work poorly on sequences requiring long-term memorization, despite having this capacity in principle. We aim to address both shortcomings with a class of recurrent networks that use a stochastic state transition mechanism between cell applications. This mechanism, which we term state-regularization, makes RNNs transition between a finite set of learnable states. We evaluate state-regularized RNNs on (1) regular languages for the purpose of automata extraction; (2) nonregular languages such as balanced parentheses, palindromes, and the copy task where external memory is required; and (3) real-word sequence learning tasks for sentiment analysis, visual object recognition, and language modeling. We show that state-regularization (a) simplifies the extraction of finite state automata modeling an RNN's state transition dynamics; (b) forces RNNs to operate more like automata with external memory and less like finite state machines; (c) makes RNNs have better interpretability and explainability.

cs.LG↗

arXiv2019

RecSys-DAN: Discriminative Adversarial Networks for Cross-Domain Recommender Systems

Data sparsity and data imbalance are practical and challenging issues in cross-domain recommender systems. This paper addresses those problems by leveraging the concepts which derive from representation learning, adversarial learning and transfer learning (particularly, domain adaptation). Although various transfer learning methods have shown promising performance in this context, our proposed novel method RecSys-DAN focuses on alleviating the cross-domain and within-domain data sparsity and data imbalance and learns transferable latent representations for users, items and their interactions. Different from existing approaches, the proposed method transfers the latent representations from a source domain to a target domain in an adversarial way. The mapping functions in the target domain are learned by playing a min-max game with an adversarial loss, aiming to generate domain indistinguishable representations for a discriminator. Four neural architectural instances of ResSys-DAN are proposed and explored. Empirical results on real-world Amazon data show that, even without using labeled data (i.e., ratings) in the target domain, RecSys-DAN achieves competitive performance as compared to the state-of-the-art supervised methods. More importantly, RecSys-DAN is highly flexible to both unimodal and multimodal scenarios, and thus it is more robust to the cold-start recommendation which is difficult for previous methods.

cs.IR↗

arXiv2019

PandaX-II Constraints on Spin-Dependent WIMP-Nucleon Effective Interactions

We present PandaX-II constraints on candidate WIMP-nucleon effective interactions involving the nucleon or WIMP spin, including, in addition to standard axial spin-dependent (SD) scattering, various couplings among vector and axial currents, magnetic and electric dipole moments, and tensor interactions. The data set corresponding to a total exposure of 54-ton-days is reanalyzed to determine constraints as a function of the WIMP mass and isospin coupling. We obtain WIMP-nucleon cross section bounds of $\rm 1.6 \times 10^{-41} cm^2$ and $\rm 9.0 \times 10^{-42} cm^2$ ($90\%$ c.l.) for neutron-only SD and tensor coupling, respectively, for a mass $M_\mathrm{WIMP} \sim {\rm 40~GeV}/c^2$. The SD limits are the best currently available for $M_\mathrm{WIMP} > {\rm 40~GeV}/c^2$. We show that PandaX-II has reached a sensitivity sufficient to probe a variety of other candidate spin-dependent interactions at the weak scale.

hep-ex↗

arXiv2019

A third order exponential time differencing numerical scheme for no-slope-selection epitaxial thin film model with energy stability

In this paper we propose and analyze a (temporally) third order accurate exponential time differencing (ETD) numerical scheme for the no-slope-selection (NSS) equation of the epitaxial thin film growth model, with Fourier pseudo-spectral discretization in space. A linear splitting is applied to the physical model, and an ETD-based multistep approximation is used for time integration of the corresponding equation. In addition, a third order accurate Douglas-Dupont regularization term, in the form of $-A \dt^2 ϕ_0 (L_N) Δ_N^2 ( u^{n+1} - u^n)$, is added in the numerical scheme. A careful Fourier eigenvalue analysis results in the energy stability in a modified version, and a theoretical justification of the coefficient $A$ becomes available. As a result of this energy stability analysis, a uniform in time bound of the numerical energy is obtained. And also, the optimal rate convergence analysis and error estimate are derived in details, in the $\ell^\infty (0,T; H_h^1) \cap \ell^2 (0,T; H_h^3)$ norm, with the help of a careful eigenvalue bound estimate, combined with the nonlinear analysis for the NSS model. This convergence estimate is the first such result for a third order accurate scheme for a gradient flow. Some numerical simulation results are presented to demonstrate the efficiency of the numerical scheme and the third order convergence. The long time simulation results for $\varepsilon=0.02$ (up to $T=3 \times 10^5$) have indicated a logarithm law for the energy decay, as well as the power laws for growth of the surface roughness and the mound width. In particular, the power index for the surface roughness and the mound width growth, created by the third order numerical scheme, is more accurate than those produced by certain second order energy stable schemes in the existing literature.

math.NA↗

arXiv2019

Ultra-low loss integrated visible photonics using thin-film lithium niobate

Integrated photonics is a powerful platform that can improve the performance and stability of optical systems, while providing low-cost, small-footprint and scalable alternatives to implementations based on free-space optics. While great progress has been made on the development of low-loss integrated photonics platforms at telecom wavelengths, visible wavelength range has received less attention. Yet, many applications utilize visible or near-visible light, including those in optical imaging, optogenetics, and quantum science and technology. Here we demonstrate an ultra-low loss integrated visible photonics platform based on thin film lithium niobate on insulator. Our waveguides feature ultra-low propagation loss of 6 dB/m, while our microring resonators have an intrinsic quality factor of 11 million, both measured at 637 nm wavelength. Additionally, we demonstrate an on-chip visible intensity modulator with an electro-optic bandwidth of 10 GHz, limited by the detector used. The ultra-low loss devices demonstrated in this work, together with the strong second- and third-order nonlinearities in lithium niobate, open up new opportunities for creating novel passive, and active devices for frequency metrology and quantum information processing in the visible spectrum range.

physics.app-ph↗

arXiv2019

An integrated low-voltage broadband lithium niobate phase modulator

Electro-optic phase modulators are critical components in modern communication, microwave photonic, and quantum photonic systems. Important for these applications is to achieve modulators with low half-wave voltage at high frequencies. Here we demonstrate an integrated phase modulator, based on a thin-film lithium niobate platform, that simultaneously features small on-chip loss (~ 1 dB) and low half-wave voltage over a large spectral range (3.5 - 4.5 V at 5 - 40 GHz). By driving the modulator with a strong 30-GHz microwave signal corresponding to around four half-wave voltages, we generate an optical frequency comb consisting of over 40 sidebands spanning 10 nm in the telecom L-band. The high electro-optic performance combined with the high RF power-handling ability (3.1 W) of our integrated phase modulator are crucial for future photonics and microwave systems.

physics.app-ph↗

arXiv2019

Low-loss fiber-to-chip interface for lithium niobate photonic integrated circuits

Integrated lithium niobate (LN) photonic circuits have recently emerged as a promising candidate for advanced photonic functions such as high-speed modulation, nonlinear frequency conversion and frequency comb generation. For practical applications, optical interfaces that feature low fiber-to-chip coupling losses are essential. So far, the fiber-to-chip loss (commonly > 10 dB) dominates the total insertion losses of typical LN photonic integrated circuits, where on-chip propagation losses can be as low as 0.03 - 0.1 dB/cm. Here we experimentally demonstrate a low-loss mode size converter for coupling between a standard lensed fiber and sub-micrometer LN rib waveguides. The coupler consists of two inverse tapers that convert the small optical mode of a rib waveguide into a symmetric guided mode of a LN nanowire, featuring a larger mode area matched to that of a tapered optical fiber. The measured fiber-to-chip coupling loss is lower than 1.7 dB/facet with high fabrication tolerance and repeatability. Our results open door for practical integrated LN photonic circuits efficiently interfaced with optical fibers.

physics.app-ph↗

arXiv2019

Positivity-preserving, energy stable numerical schemes for the Cahn-Hilliard equation with logarithmic potential

We present and analyze finite difference numerical schemes for the Allen Cahn/Cahn-Hilliard equation with a logarithmic Flory Huggins energy potential. Both the first order and second order accurate temporal algorithms are considered. In the first order scheme, we treat the nonlinear logarithmic terms and the surface diffusion term implicitly, and update the linear expansive term and the mobility explicitly. We provide a theoretical justification that, this numerical algorithm has a unique solution such that the positivity is always preserved for the logarithmic arguments. In particular, our analysis reveals a subtle fact: the singular nature of the logarithmic term around the values of $-1$ and 1 prevents the numerical solution reaching these singular values, so that the numerical scheme is always well-defined as long as the numerical solution stays similarly bounded at the previous time step. Furthermore, an unconditional energy stability of the numerical scheme is derived, without any restriction for the time step size. The unique solvability and the positivity-preserving property for the second order scheme are proved using similar ideas, in which the singular nature of the logarithmic term plays an essential role. For both the first and second order accurate schemes, we are able to derive an optimal rate convergence analysis, which gives the full order error estimate. The case with a non-constant mobility is analyzed as well. We also describe a practical and efficient multigrid solver for the proposed numerical schemes, and present some numerical results, which demonstrate the robustness of the numerical schemes.

math.NA↗

arXiv2019

Penalized Interaction Estimation for Ultrahigh Dimensional Quadratic Regression

Quadratic regression goes beyond the linear model by simultaneously including main effects and interactions between the covariates. The problem of interaction estimation in high dimensional quadratic regression has received extensive attention in the past decade. In this article we introduce a novel method which allows us to estimate the main effects and interactions separately. Unlike existing methods for ultrahigh dimensional quadratic regressions, our proposal does not require the widely used heredity assumption. In addition, our proposed estimates have explicit formulas and obey the invariance principle at the population level. We estimate the interactions of matrix form under penalized convex loss function. The resulting estimates are shown to be consistent even when the covariate dimension is an exponential order of the sample size. We develop an efficient ADMM algorithm to implement the penalized estimation. This ADMM algorithm fully explores the cheap computational cost of matrix multiplication and is much more efficient than existing penalized methods such as all pairs LASSO. We demonstrate the promising performance of our proposal through extensive numerical studies.

stat.ME↗

arXiv2019

A Fast Iterative Algorithm for High-dimensional Differential Network

Differential network is an important tool to capture the changes of conditional correlations under two sample cases. In this paper, we introduce a fast iterative algorithm to recover the differential network for high-dimensional data. The computation complexity of our algorithm is linear in the sample size and the number of parameters, which is optimal in the sense that it is of the same order as computing two sample covariance matrices. The proposed method is appealing for high-dimensional data with a small sample size. The experiments on simulated and real data sets show that the proposed algorithm outperforms other existing methods.

stat.CO↗

arXiv2019

Towards Personalized Management of Type B Aortic Dissection Using STENT: a STandard cta database with annotation of the ENtire aorta and True-false lumen

Type B Aortic Dissection(TBAD) is a rare aortic disease with a high 5-year mortality.Personalized and precise management of TBAD has been increasingly desired in clinic which requires the geometric parameters of TBAD specific to the patient be measured accurately.This remains to be a challenging task for vascular surgeons as manual measurement is highly subjective and imprecise. To solve this problem,we introduce STENT-a STandard cta database with annotation of the ENtire aorta and True-false lumen. The database contains 274 CT angiography (CTA) scans from 274 unique TBAD patients and is split into a training set(254 cases including 210 preoperative and 44 postoperative scans ) and a test set(20 cases).Based on STENT,we develop a series of methods including automated TBAD segmentation and automated measurement of TBAD parameters that facilitate personalized and precise management of the disease. In this work, the database and the proposed methods are thoroughly introduced and evaluated and the results of our study shows the feasibility and effectiveness of our approach to easing the decision-making process for vascular surgeons during personalized TBAD management.

physics.med-ph↗

arXiv2019

Integrating and querying similar tables from PDF documents using deep learning

Large amount of public data produced by enterprises are in semi-structured PDF form. Tabular data extraction from reports and other published data in PDF format is of interest for various data consolidation purposes such as analysing and aggregating financial reports of a company. Queries into the structured tabular data in PDF format are normally processed in an unstructured manner through means like text-match. This is mainly due to that the binary format of PDF documents is optimized for layout and rendering and do not have great support for automated parsing of data. Moreover, even the same table type in PDF files varies in schema, row or column headers, which makes it difficult for a query plan to cover all relevant tables. This paper proposes a deep learning based method to enable SQL-like query and analysis of financial tables from annual reports in PDF format. This is achieved through table type classification and nearest row search. We demonstrate that using word embedding trained on Google news for header match clearly outperforms the text-match based approach in traditional database. We also introduce a practical system that uses this technology to query and analyse finance tables in PDF documents from various sources.

cs.IR↗

arXiv2019

Multi-level Chirality in Liquid Crystals Formed by Achiral Molecules

In many biological materials with a hierarchical structure there is an intriguing and unique mechanism responsible for the 'propagation' of order from the molecular to the nano- or micro-scale level. Here we present a much simpler molecular system built of achiral mesogenic dimeric molecules that shows a similar complexity with four levels of structural chirality (i) layer chirality, (ii) helicity of a basic 4-layer repeating unit, (iii) a helix with a pitch of several layers and (iv) mesoscopic helical filaments. As seen in many biological systems, there is a coupling between chirality at different levels. The structures were identified by a combination of hard and soft x-ray diffraction measurements, optical studies and theoretical modelling.

cond-mat.soft↗

arXiv2018

Geometric Multi-Model Fitting by Deep Reinforcement Learning

This paper deals with the geometric multi-model fitting from noisy, unstructured point set data (e.g., laser scanned point clouds). We formulate multi-model fitting problem as a sequential decision making process. We then use a deep reinforcement learning algorithm to learn the optimal decisions towards the best fitting result. In this paper, we have compared our method against the state-of-the-art on simulated data. The results demonstrated that our approach significantly reduced the number of fitting iterations.

cs.CV↗

arXiv2018

NeuroTreeNet: A New Method to Explore Horizontal Expansion Network

It is widely recognized that the deeper networks or networks with more feature maps have better performance. Existing studies mainly focus on extending the network depth and increasing the feature maps of networks. At the same time, horizontal expansion network (e.g. Inception Model) as an alternative way to improve network performance has not been fully investigated. Accordingly, we proposed NeuroTreeNet (NTN), as a new horizontal extension network through the combination of random forest and Inception Model. Based on the tree structure, in which each branch represents a network and the root node features are shared to child nodes, network parameters are effectively reduced. By combining all features of leaf nodes, even less feature maps achieved better performance. In addition, the relationship between tree structure and the performance of NTN was investigated in depth. Comparing to other networks (e.g. VDSR\_5) with equal magnitude parameters, our model showed preferable performance in super resolution reconstruction task.

cs.CV↗

arXiv2018

Estimation of Spatial-Temporal Gait Parameters based on the Fusion of Inertial and Film-Pressure Signals

Gait analysis (GA) has been widely used in physical activity monitoring and clinical contexts, and the estimation of the spatial-temporal gait parameters is of primary importance for GA. With the quick development of smart tiny sensors, GA methods based on wearable devices have become more popular recently. However, most existing wearable GA methods focus on data analysis from inertial sensors. In this paper, we firstly present a two-foot-worn in-shoe system (Gaitboter) based on low-cost, wearable and multimodal sensors' fusion for GA, comprising an inertial sensor and eight film-pressure sensors with each foot for gait raw data collection while walking. Secondly, a GA algorithm for estimating the spatial-temporal parameters of gait is proposed. The algorithm fully uses the fusion of two kinds of sensors' signals: inertial sensor and film-pressure sensor, in order to estimate the spatial-temporal gait parameters, such as stance phase, swing phase, double stance phase, cadence, step time, stride time, stride length, velocity. Finally, to verify the effectiveness of this system and algorithm of the paper, an experiment is performed with 27 stoke patients from local hospital that the spatial-temporal gait parameters obtained with the system and the algorithm are compared with a GA tool used in medical laboratories. And the experimental results show that it achieves very good consistency and correlation between the proposed system and the compared GA tool.

eess.SP↗

arXiv2018

Influence of chemical kinetics on detonation initiating by temperature gradients in methane/air

Different simplified and detailed chemical models and their impact on simulations of combustion regimes initiating by the initial temperature gradient in methane/air mixtures are studied. The limits of the regimes of reaction wave propagation depend upon the spontaneous wave speed and the characteristic velocities of the problem. The present study mainly focus to identify conditions required for the development a detonation and to compare the difference between simplified chemical models and detailed chemistry. It is shown that a widely used simplified chemical schemes, such as one-step, two-step and other simplified models, do not reproduce correctly the ignition process in methane/air mixtures. The ignition delay times calculated using simplified models are in orders of magnitude shorter than the ignition delay times calculated using detailed chemical models and measured experimentally. This results in considerably different times when the exothermic reaction affects significantly the ignition, evolution, and coupling of the spontaneous reaction wave and pressure waves. We show that the temperature gradient capable to trigger detonation calculated using detailed chemical models is much shallower (the size of the hot spot is much larger) than that, predicted by simulations with simplified chemical models. These findings suggest that the scenario leading to the deflagration to detonation transition (DDT) may depend greatly on the chemical model used in simulations and that the Zeldovich gradient mechanism is not necessary a universal mechanism triggering DDT. The obtained results indicate that the conclusions derived from the simulations of DDT with simplified chemical models should be viewed with great caution.

physics.chem-ph↗

arXiv2018

Ultrahigh-efficiency second-harmonic generation in nanophotonic PPLN waveguides

Periodically poled lithium niobate (PPLN) waveguide is a powerful platform for efficient wavelength conversion. Conventional PPLN converters however typically require long device lengths and high pump powers due to the limited nonlinear interaction strength. Here we use a nanostructured PPLN waveguides to demonstrate an ultrahigh normalized efficiency of 2600%/W-cm$^2$ for second-harmonic generation of 1.5-$μ$m radiation, more than 20 times higher than that in state-of-the-art diffused waveguides. This is achieved by a combination of sub-wavelength optical confinement and high-fidelity periodic poling at a first-order poling period of 4 $μ$m. Our highly integrated PPLN waveguides are promising for future chip-scale integration of classical and quantum photonic systems.

physics.app-ph↗