arXiv subjects

Yan Li

Publications and source records attributed to Yan Li.

At least 577 records · Page 32Linked to original sources

arXiv2018

M$^2$S-Net: Multi-Modal Similarity Metric Learning based Deep Convolutional Network for Answer Selection

Recent works using artificial neural networks based on distributed word representation greatly boost performance on various natural language processing tasks, especially the answer selection problem. Nevertheless, most of the previous works used deep learning methods (like LSTM-RNN, CNN, etc.) only to capture semantic representation of each sentence separately, without considering the interdependence between each other. In this paper, we propose a novel end-to-end learning framework which constitutes deep convolutional neural network based on multi-modal similarity metric learning (M$^2$S-Net) on pairwise tokens. The proposed model demonstrates its performance by surpassing previous state-of-the-art systems on the answer selection benchmark, i.e., TREC-QA dataset, in both MAP and MRR metrics.

cs.CL↗

arXiv2018

All-optical quantum signal demultiplexer

Dense wavelength division multiplexing (DWDM) is one of the most successful methods for enhancing data transmission rates in both classical and quantum communication networks. Although signal multiplexing and demultiplexing are equally important, traditional multiplexing and demultiplexing methods are based on passive devices such as arrayed waveguides and fiber Bragg cascade filters, which, although widely used in commercial devices, lack any active tuning ability. In this work, we propose a signal demultiplexing method based on sum frequency generation (SFG) with two significant features: first, any signal from the common communication channel can be demultiplexed to a single user by switching the pump wavelength; second, a cheap high-performance detector can be used for signal detection. These two features were demonstrated by demultiplexing multi-channel energy-time entanglement generated by a micro-cavity silicon chip. High interference visibilities over three channels after demultiplexing showed that entanglement was preserved and verified the high performance of the demultiplexer, which will find wide application in high-capacity quantum communication networks.

quant-ph↗

arXiv2018

A Framework in CRM Customer Lifecycle: Identify Downward Trend and Potential Issues Detection

Customer retention is one of the primary goals in the area of customer relationship management. A mass of work exists in which machine learning models or business rules are established to predict churn. However, targeting users at an early stage when they start to show a downward trend is a better strategy. In downward trend prediction, the reasons why customers show a downward trend is of great interest in the industry as it helps the business to understand the pain points that customers suffer and to take early action to prevent them from churning. A commonly used method is to collect feedback from customers by either aggressively reaching out to them or by passively hearing from them. However, it is believed that there are a large number of customers who have unpleasant experiences and never speak out. In the literature, there is limited research work that provides a comprehensive and scientific approach to identify these "silent suffers". In this study, we propose a novel two-part framework: developing the downward prediction process and establishing the methodology to identify the reasons why customers are in the downward trend. In the first prediction part, we focus on predicting the downward trend, which is an earlier stage of the customer lifecycle compared to churn. In the second part, we propose an approach to figuring out the cause (of the downward trend) based on a causal inference method and semi-supervised learning. The proposed approach is capable of identifying potential silent sufferers. We take bad shopping experiences as inputs to develop the framework and validate it via a marketing A/B test in the real world. The test readout demonstrates the effectiveness of the framework by driving 88.5% incremental lift in purchase volume.

cs.CY↗

arXiv2018

Fast Global Convergence via Landscape of Empirical Loss

While optimizing convex objective (loss) functions has been a powerhouse for machine learning for at least two decades, non-convex loss functions have attracted fast growing interests recently, due to many desirable properties such as superior robustness and classification accuracy, compared with their convex counterparts. The main obstacle for non-convex estimators is that it is in general intractable to find the optimal solution. In this paper, we study the computational issues for some non-convex M-estimators. In particular, we show that the stochastic variance reduction methods converge to the global optimal with linear rate, by exploiting the statistical property of the population loss. En route, we improve the convergence analysis for the batch gradient method in \cite{mei2016landscape}.

stat.ML↗

arXiv2018

Frequency identification and asteroseismic analysis of the red giant KIC 9145955: fundamental parameters and helium core size

We have analyzed 18 quarters of long-cadence data of KIC 9145955 provided by \emph{Kepler}, and extracted 61 oscillation frequencies from these high precision photometric data. The oscillation frequencies include 7 $l = 0$ modes, 44 $l = 1$ modes, 7 $l = 2$ modes, and 3 $l = 3$ modes. We identify $l = 0$ modes as p modes and $l = 2$ modes as p-dominated modes. For $l = 1$ modes, all of them are identified as mixed modes. These mixed modes can be used to determine the size of the helium core. We conduct a series of asteroseismic models and the size of the helium core is determined to be $M_{\rm He}$ = 0.210 $\pm$ 0.002 $M_{\odot}$ and $R_{\rm He}$ = 0.0307 $\pm$ 0.0002 $R_{\odot}$. Furthermore, we find that only the acoustic radius $τ_{0}$ can be precisely determined with the asteroseismic method independently. The value of $τ_{0}$ is determined to be 0.494 $\pm$ 0.001 days. By combining asteroseismic results and spectroscopic observations, we obtain the best-fitting model. The physical parameters of this model are $M$ = 1.24 $M_{\odot}$, $Z$ = 0.009, $α$ = 2.0, $T_{\rm eff}$ = 5069 K, $\log g$ = 3.029, $R$ = 5.636 $R_{\odot}$, and $L$ = 18.759 $L_{\odot}$. In addition, we think that the observed frequency F39 (96.397 $μ$Hz) is more appropriate to be identified as a mixed mode of the most p-dominated.

astro-ph.SR↗

arXiv2018

A generalization of Menon's identity with Dirichlet characters

The classical Menon's identity [7] states that \begin{equation*}\label{oldbegin1} \sum_{\substack{a\in\Bbb Z_n^\ast }}\gcd(a -1,n)=φ(n) σ_{0} (n), \end{equation*} where for a positive integer $n$, $\Bbb Z_n^\ast$ is the group of units of the ring $\Bbb Z_n=\Bbb Z/n\Bbb Z$, $\gcd(\ ,\ )$ represents the greatest common divisor, $φ(n)$ is the Euler's totient function and $σ_{k} (n) =\sum_{d|n } d^{k}$ is the divisor function. In this paper, we generalize Menon's identity with Dirichlet characters in the following way: \begin{equation*} \sum_{\substack{a\in\Bbb Z_n^\ast b_1, ..., b_k\in\Bbb Z_n}} \gcd(a-1,b_1, ..., b_k, n)χ(a)=φ(n)σ_k\left(\frac{n}{d}\right), \end{equation*} where $k$ is a non-negative integer and $χ$ is a Dirichlet character modulo $n$ whose conductor is $d$. Our result can be viewed as an extension of Zhao and Cao's result [16] to $k>0$. It can also be viewed as an extension of Sury's result [12] to Dirichlet characters.

math.NT↗

arXiv2018

$G$-Sasaki manifolds and K-energy

In this paper, we introduce a class of Sasaki manifolds with a reductive $G$-group action, called $G$-Sasaki manifolds. By reducing K-energy to a functional defined on a class of convex functions on a moment polytope, we give a criterion for the properness of K-energy. In particular, we deduce a sufficient and necessary condition related to the polytope for the existence of transverse $G$-Sasaki Einstein metrics. A similar result is also obtained for transverse $G$-Sasaki Ricci solitons. As an application, we construct several examples of $G$-Sasaki Ricci solitons by an established openness theorem for transverse $G$-Sasaki Ricci solitons.

math.DG↗

arXiv2018

Quantum oscillations and nontrivial transport in (Bi0.92In0.08)2Se3

Quantum phase transition in topological insulators has drawn heightened attention in condensed matter physics and future device applications. Here we report the magnetotransport properties of single crystalline (Bi0.92In0.08)2Se3. The average mobility of about 1000 cm2/Vs is obtained from the Lorentz law at the low field up to 50 K. The quantum oscillations rise at a field of about 5 T, revealing a high mobility of 1.4*10^4 cm2/Vs at 2 K. The topological Dirac fermions are evident by the nontrivial Berry phase in the Landau Fan diagram. The properties make the (Bi0.92In0.08)2Se3 a promising platform for the investigation of quantum phase transition in topological insulators.

cond-mat.mtrl-sci↗

arXiv2017

Generation of Bose-Einstein Condensates' Ground State Through Machine Learning

We show that both single-component and two-component Bose-Einstein condensates' (BECs) ground states can be simulated by deep convolutional neural networks of the same structure. We trained the neural network via inputting the coupling strength in the dimensionless Gross-Pitaevskii equation (GPE) and outputting the ground state wave-function. After training, the neural network generates ground states faster than the method of imaginary time evolution, while the relative mean-square-error between predicted states and original states is in the magnitude between $10^{-5}$ and $10^{-4}$. We compared the eigen-energies based on predicted states and original states, it is shown that the neural network can predict eigen-energies in high precisions. Therefore, the BEC ground states, which are continuous wave-functions, can be represented by deep convolution neural networks.

quant-ph↗

arXiv2017

Approximate dispersion relations for waves on arbitrary shear flows

An approximate dispersion relation is derived and presented for linear surface waves atop a shear current whose magnitude and direction can vary arbitrarily with depth. The approximation, derived to first order of deviation from potential flow, is shown to produce good approximations at all wavelengths for a wide range of naturally occuring shear flows as well as widely used model flows. The relation reduces in many cases to a 3D generalization of the much used approximation by Skop [1987], developed further by Kirby & Chen [1989], but is shown to be more robust, succeeding in situations where the Kirby & Chen model fails. The two approximations incur the same numerical cost and difficulty. While the Kirby & Chen approximation is excellent for a wide range of currents, the exact criteria for its applicability have not been known. We explain the apparently serendipitous success of the latter and derive proper conditions of applicability for both approximate dispersion relations. Our new model has a greater range of applicability. A second order approximation is also derived. It greatly improves accuracy, which is shown to be important in difficult cases. It has an advantage over the corresponding 2nd order expression proposed by Kirby \& Chen that its criterion of accuracy is explicitly known, which is not currently the case for the latter to our knowledge. Our 2nd order term is also arguably significantly simpler to implement, and more physically transparent, than its sibling due to Kirby & Chen.

physics.flu-dyn↗

arXiv2017

A Hopf's Lemma and the Boundary Regularity for the Fractional P-Laplacian

We begin the paper with a Hopf's lemma for a fractional p-Laplacian problem on a half-space. Specifically speaking, we show that the derivative of the solution along the outward normal vector is strictly positive on the boundary of the half-space. Next we show that positive solutions to a fractional p-Laplacian equation possess certain Holder continuity up to the boundary.

math.AP↗

arXiv2017

Mabuchi metrics and properness of the modified DING functional

In this paper, we study Mabuchi metrics on Fano manifolds. We prove that Mabuchi metrics exist if the modified Ding functional is proper modulo a reductive subgroup of its automorphism group. On the other hand, the inverse that Mabuchi metrics implies the properness is obtained by using Darvas-Rubinstein's properness principle. As an application, we establish a criterion for the existence of Mabuchi metrics on Fano group compactifications.

math.DG↗

arXiv2017

Transient wave resistance upon a real shear current

We study the waves and wave-making forces acting on ships travelling on currents which vary as a function of depth. Our concern is realism; we consider a real current profile from the Columbia River, and model ships with dimensions and Froude numbers typical of three classes of vessels operating in these waters. To this end we employ the most general theory of waves from free-surface sources on shear current to date, which we derive and present here. Expressions are derived for ship waves which satisfy an arbitrary dispersion relation and are generated by a wave source acting on the free surface, with the source's shape and time-dependence is also being arbitrary. Practical calculation procedures for numerically calculating dispersion on a shear current which may vary arbitrarily with depth both in direction and magnitude, are indicated. For ships travelling at oblique angle to a shear-current, the ship wave pattern is asymmetrical, and wave-making radiation forces have a lateral component in addition to the conventional wave resistance, the sternward component. No corresponding lateral force exists in the absence of shear. We consider the dependence of wave resistance and lateral force for upstream, downstream and cross-stream motion on the Columbia River current, both in steady motion and during two different maneouvres: a ship suddenly set in motion, and a ship turning through 360 deg. We find that for smaller ships (tugboats, fishing-boats) the wave resistance can differ drastically from that in quiescent water, and depends strongly on Froude number and direction of motion. For Froude numbers typical of such boats, wave resistance can vary by a factor 3 between upstream and downstream motion, and the strong Froude number dependence is made more complicated by interference effects. The lateral radiation force ... [abstract truncated due to ArXiV's space restrictions]

physics.flu-dyn↗

arXiv2017

Non-convex Conditional Gradient Sliding

We investigate a projection free method, namely conditional gradient sliding on batched, stochastic and finite-sum non-convex problem. CGS is a smart combination of Nesterov's accelerated gradient method and Frank-Wolfe (FW) method, and outperforms FW in the convex setting by saving gradient computations. However, the study of CGS in the non-convex setting is limited. In this paper, we propose the non-convex conditional gradient sliding (NCGS) which surpasses the non-convex Frank-Wolfe method in batched, stochastic and finite-sum setting.

math.OC↗

arXiv2017

Machine Learning for Survival Analysis: A Survey

Accurately predicting the time of occurrence of an event of interest is a critical problem in longitudinal data analysis. One of the main challenges in this context is the presence of instances whose event outcomes become unobservable after a certain time point or when some instances do not experience any event during the monitoring period. Such a phenomenon is called censoring which can be effectively handled using survival analysis techniques. Traditionally, statistical approaches have been widely developed in the literature to overcome this censoring issue. In addition, many machine learning algorithms are adapted to effectively handle survival data and tackle other challenging problems that arise in real-world data. In this survey, we provide a comprehensive and structured review of the representative statistical methods along with the machine learning techniques used in survival analysis and provide a detailed taxonomy of the existing methods. We also discuss several topics that are closely related to survival analysis and illustrate several successful applications in various real-world application domains. We hope that this paper will provide a more thorough understanding of the recent advances in survival analysis and offer some guidelines on applying these approaches to solve new problems that arise in applications with censored data.

cs.LG↗

arXiv2017

Electrically-induced polarization selection rules of a graphene quantum dot

We study theoretically the single-electron triangular zigzag graphene quantum dot in uniform in-plane electric fields. The far-infrared absorption spectra of the dot are calculated by the tight-binding method. The energy spectra and the distribution of wave functions are also presented to analyse the far-infrared spectra. The orthogonal zero-energy eigenstates are arranged along to the direction of the external field. The remarkable result is that all intraband transitions and some interband transitions are forbidden when the absorbed light is polarized along the direction of the electric field. With x-direction electric field, all intraband absorption is y polarized due to the electric-field-direction-polarization selection rule. Moreover, with y-direction electric field, all absorption is either x or y polarized due to the parity selection rule as well as to the electric-field-direction-polarization selection rule. Our calculation shows that the formation of the FIR spectra is co-decided by the polarization selection rules and the overlap between the eigenstates of the transition.

cond-mat.mes-hall↗

arXiv2017

In-situ measurement of light polarization with ellipticity-induced nonlinear magneto-optical rotation

A precise, accurate, and relatively straightforward in-situ method to measure and control the ellipticity of light resonantly interacting with an atomic vapor is described. The technique can be used to minimize vector light shifts. The method involves measurement of ellipticity-induced resonances in the magnetic-field dependence of nonlinear magneto-optical rotation of frequency-modulated light. The light propagation direction is orthogonal to the applied magnetic field $\textbf{B}$ and the major axis of the light polarization ellipse is along $\textbf{B}$. When the light modulation frequency matches the Larmor frequency, elliptically polarized light produces precessing atomic spin orientation transverse to $\textbf{B}$ via synchronous optical pumping. The precessing spin orientation causes optical rotation oscillating at the Larmor frequency by modulating the atomic vapor's circular birefringence. Based on this technique's precision, in-situ nature (which avoids systematic errors arising from optical interfaces), and independent control of the most important systematic errors, it is shown that the accuracy of light ellipticity measurements achievable with this technique can exceed that of existing methods by orders of magnitude.

physics.optics↗

arXiv2017

Linear Convergence of SVRG in Statistical Estimation

SVRG and its variants are among the state of art optimization algorithms for large scale machine learning problems. It is well known that SVRG converges linearly when the objective function is strongly convex. However this setup can be restrictive, and does not include several important formulations such as Lasso, group Lasso, logistic regression, and some non-convex models including corrected Lasso and SCAD. In this paper, we prove that, for a class of statistical M-estimators covering examples mentioned above, SVRG solves the formulation with {\em a linear convergence rate} without strong convexity or even convexity. Our analysis makes use of {\em restricted strong convexity}, under which we show that SVRG converges linearly to the fundamental statistical precision of the model, i.e., the difference between true unknown parameter $θ^*$ and the optimal solution $\hatθ$ of the model.

stat.ML↗