arXiv subjects

Pengfei Li

Publications and source records attributed to Pengfei Li.

At least 235 records · Page 13Linked to original sources

arXiv2020

Semiparametric inference on general functionals of two semicontinuous populations

In this paper, we propose new semiparametric procedures for making inference on linear functionals and their functions of two semicontinuous populations. The distribution of each population is usually characterized by a mixture of a discrete point mass at zero and a continuous skewed positive component, and hence such distribution is semicontinuous in the nature. To utilize the information from both populations, we model the positive components of the two mixture distributions via a semiparametric density ratio model. Under this model setup, we construct the maximum empirical likelihood estimators of the linear functionals and their functions, and establish the asymptotic normality of the proposed estimators. We show the proposed estimators of the linear functionals are more efficient than the fully nonparametric ones. The developed asymptotic results enable us to construct confidence regions and perform hypothesis tests for the linear functionals and their functions. We further apply these results to several important summary quantities such as the moments, the mean ratio, the coefficient of variation, and the generalized entropy class of inequality measures. Simulation studies demonstrate the advantages of our proposed semiparametric method over some existing methods. Two real data examples are provided for illustration.

stat.ME↗

arXiv2020

On Convergence of Gradient Expected Sarsa($λ$)

We study the convergence of $\mathtt{Expected~Sarsa}(λ)$ with linear function approximation. We show that applying the off-line estimate (multi-step bootstrapping) to $\mathtt{Expected~Sarsa}(λ)$ is unstable for off-policy learning. Furthermore, based on convex-concave saddle-point framework, we propose a convergent $\mathtt{Gradient~Expected~Sarsa}(λ)$ ($\mathtt{GES}(λ)$) algorithm. The theoretical analysis shows that our $\mathtt{GES}(λ)$ converges to the optimal solution at a linear convergence rate, which is comparable to extensive existing state-of-the-art gradient temporal difference learning algorithms. Furthermore, we develop a Lyapunov function technique to investigate how the step-size influences finite-time performance of $\mathtt{GES}(λ)$, such technique of Lyapunov function can be potentially generalized to other GTD algorithms. Finally, we conduct experiments to verify the effectiveness of our $\mathtt{GES}(λ)$.

cs.LG↗

arXiv2020

Metastable soliton necklaces supported by fractional diffraction and competing nonlinearities

We demonstrate that fractional cubic-quintic nonlinear Schrödinger equation,characterized by its Lévy index, maintains ring-shaped soliton clusters ("necklaces") carrying orbital angular momentum. They can be built, in the respective optical setting, as circular chains of fundamental solitons linked by a vortical phase field. We predict semi-analytically that the metastable necklace-shaped clusters persist, corresponding to a local minimum of an effective potential of interaction between adjacent solitons in the cluster. Systematic simulations corroborate that the clusters stay robust over extremely large propagation distances, even in the presence of strong random perturbations.

nlin.PS↗

arXiv2020

Characteristic mass in galaxy quenching: environmental versus internal effects

A clear transition feature of galaxy quenching is identified in the multi-parameter space of stellar mass ($M_*$), bulge to total mass ratio ($B/T_{\rm m}$), halo mass ($M_{\rm h}$) and halo-centric distance ($r/r_{180}$). For given halo mass, the characteristic stellar mass ($M_{*, \rm ch}$) for the transition is about one-fifth of that of the corresponding central galaxy, and almost independent of $B/T_{\rm m}$. Once $B/T_{\rm m}$ is fixed, the quenched fraction of galaxies with $M_{*} < M_{*, \rm ch}$ increases with $M_{\rm h}$, but decreases with $M_*$ in the inner part of halos ($r/r_{180} < 0.5$). In the outer part ($r/r_{180} > 0.5$), the trend with $M_{\rm h}$ remains but the correlation with $M_*$ is absent or becomes positive. For galaxies above $M_{\rm *, ch}$ and with $B/T_{\rm m}$ fixed, the quenched fraction increases with $M_{\rm *}$, but depends only weakly on $M_{\rm h}$ in both the inner and outer regions. At fixed $B/T_{\rm m}$ and $M_*$, the quenched fraction increases with decreasing $r/r_{180}$ for galaxies with $M_{*} < M_{*, \rm ch}$, and depends only weakly on $r/r_{180}$ for galaxies with $M_{*} > M_{*, \rm ch}$. Our finding provides a physically-motivated way to classify galaxies in halos into two classes based on their quenching properties: an `upper class' with $M_{*} > M_{\rm *,ch}$ and a `lower class' with $M_{*} < M_{\rm *,ch}$. Environmental quenching is important for `lower class' galaxies, while internal quenching plays the dominating role for the `upper class'.

astro-ph.GA↗

arXiv2020

Multi-person 3D Pose Estimation in Crowded Scenes Based on Multi-View Geometry

Epipolar constraints are at the core of feature matching and depth estimation in current multi-person multi-camera 3D human pose estimation methods. Despite the satisfactory performance of this formulation in sparser crowd scenes, its effectiveness is frequently challenged under denser crowd circumstances mainly due to two sources of ambiguity. The first is the mismatch of human joints resulting from the simple cues provided by the Euclidean distances between joints and epipolar lines. The second is the lack of robustness from the naive formulation of the problem as a least squares minimization. In this paper, we depart from the multi-person 3D pose estimation formulation, and instead reformulate it as crowd pose estimation. Our method consists of two key components: a graph model for fast cross-view matching, and a maximum a posteriori (MAP) estimator for the reconstruction of the 3D human poses. We demonstrate the effectiveness and superiority of our proposed method on four benchmark datasets.

cs.CV↗

arXiv2020

Semiparametric Inference of the Youden Index and the Optimal Cutoff Point under Density Ratio Models

The Youden index is a popular summary statistic for receiver operating characteristic curve. It gives the optimal cutoff point of a biomarker to distinguish the diseased and healthy individuals. In this paper, we propose to model the distributions of a biomarker for individuals in the healthy and diseased groups via a semiparametric density ratio model. Based on this model, we use the maximum empirical likelihood method to estimate the Youden index and the optimal cutoff point. We further establish the asymptotic normality of the proposed estimators and construct valid confidence intervals for the Youden index and the corresponding optimal cutoff point. The proposed method automatically covers both cases when there is no lower limit of detection (LLOD) and when there is a fixed and finite LLOD for the biomarker. Extensive simulation studies and a real data example are used to illustrate the effectiveness of the proposed method and its advantages over the existing methods.

stat.ME↗

arXiv2020

Vortex solitons in fractional nonlinear Schrödinger equation with the cubic-quintic nonlinearity

We address the existence and stability of vortex-soliton (VS) solutions of the fractional nonlinear Schrödinger equation (NLSE) with competing cubic-quintic nonlinearities and the Lévy index (fractionality) taking values 1 \leqα\leq2. Families of ring-shaped VSs with vorticities s = 1,2, and 3 are constructed in a numerical form. Unlike the usual two-dimensional NLSE (which corresponds to α = 2), in the fractional model VSs exist above a finite threshold value of the total power,P. Stability of the VS solutions is investigated for small perturbations governed by the linearized equation, and corroborated by direct simulations. Unstable VSs are broken up by azimuthal perturbations into several fragments, whose number is determined by the fastest growing eigenmode of small perturbations. The stability region, defined in terms of P, expands with the increase of α from 1 up to 2 for all s = 1, 2, and 3, except for steep shrinkage for s = 2 in the interval of 1\leqα\leq1.3.

nlin.PS↗

arXiv2020

Symmetry breaking of spatial Kerr solitons in fractional dimension

We study symmetry breaking of solitons in the framework of a nonlinear fractional Schrödinger equation (NLFSE), characterized by its Lévy index, with cubic nonlinearity and a symmetric double-well potential. Asymmetric, symmetric, and antisymmetric soliton solutions are found, with stable asymmetric soliton solutions emerging from unstable symmetric and antisymmetric ones by way of symmetry-breaking bifurcations. Two different bifurcation scenarios are possible. First, symmetric soliton solutions undergo a symmetry-breaking bifurcation of the pitchfork type, which gives rise to a branch of asymmetric solitons, under the action of the self-focusing nonlinearity. Second, a family of asymmetric solutions branches off from antisymmetric states in the case of self-defocusing nonlinearity through a bifurcation of an inverted-pitchfork type. Systematic numerical analysis demonstrates that increase of the Lévy index leads to shrinkage or expansion of the symmetry-breaking region, depending on parameters of the double-well potential. Stability of the soliton solutions is explored following the variation of the Lévy index, and the results are confirmed by direct numerical simulations.

nlin.PS↗

arXiv2020

Improving Relation Extraction with Knowledge-attention

While attention mechanisms have been proven to be effective in many NLP tasks, majority of them are data-driven. We propose a novel knowledge-attention encoder which incorporates prior knowledge from external lexical resources into deep neural networks for relation extraction task. Furthermore, we present three effective ways of integrating knowledge-attention with self-attention to maximize the utilization of both knowledge and data. The proposed relation extraction system is end-to-end and fully attention-based. Experiment results show that the proposed knowledge-attention mechanism has complementary strengths with self-attention, and our integrated models outperform existing CNN, RNN, and self-attention based models. State-of-the-art performance is achieved on TACRED, a complex and large-scale relation extraction dataset.

cs.CL↗

arXiv2020

A comprehensive catalog of dark matter halo models for SPARC galaxies

We present rotation curve fits to 175 late-type galaxies from the Spitzer Photometry & Accurate Rotation Curves (SPARC) database using seven dark matter (DM) halo profiles: pseudo-isothermal (pISO), Burkert, Navarro-Frenk-White (NFW), Einasto, Di Cintio (2014, DC14), coreNFW, and a new semi-empirical profile named Lucky13. We marginalize over stellar mass-to-light ratio, galaxy distance, disk inclination, halo concentration and halo mass (and an additional shape parameter for Einasto) using a Markov Chain Monte Carlo method. We find that cored halo models such as the DC14 and Burkert profiles generally provide better fits to rotation curves than the cuspy NFW profile. The stellar mass-halo mass relation from abundance matching is recovered by all halo profiles once imposed as a Bayesian prior, whereas the halo mass-concentration relation is not reproduced in detail by any halo model. We provide an extensive set of figures as well as best-fit parameters in machine-readable tables to facilitate model comparison and the exploration of DM halo properties.

astro-ph.GA↗

arXiv2019

Car Pose in Context: Accurate Pose Estimation with Ground Plane Constraints

Scene context is a powerful constraint on the geometry of objects within the scene in cases, such as surveillance, where the camera geometry is unknown and image quality may be poor. In this paper, we describe a method for estimating the pose of cars in a scene jointly with the ground plane that supports them. We formulate this as a joint optimization that accounts for varying car shape using a statistical atlas, and which simultaneously computes geometry and internal camera parameters. We demonstrate that this method produces significant improvements for car pose estimation, and we show that the resulting 3D geometry, when computed over a video sequence, makes it possible to improve on state of the art classification of car behavior. We also show that introducing the planar constraint allows us to estimate camera focal length in a reliable manner.

cs.CV↗

arXiv2019

Qualitative Measurements of Policy Discrepancy for Return-Based Deep Q-Network

The deep Q-network (DQN) and return-based reinforcement learning are two promising algorithms proposed in recent years. DQN brings advances to complex sequential decision problems, while return-based algorithms have advantages in making use of sample trajectories. In this paper, we propose a general framework to combine DQN and most of the return-based reinforcement learning algorithms, named R-DQN. We show the performance of traditional DQN can be improved effectively by introducing return-based reinforcement learning. In order to further improve the R-DQN, we design a strategy with two measurements which can qualitatively measure the policy discrepancy. Moreover, we give the two measurements' bounds in the proposed R-DQN framework. We show that algorithms with our strategy can accurately express the trace coefficient and achieve a better approximation to return. The experiments, conducted on several representative tasks from the OpenAI Gym library, validate the effectiveness of the proposed measurements. The results also show that the algorithms with our strategy outperform the state-of-the-art methods.

cs.LG↗

arXiv2019

The halo mass function of late-type galaxies from HI kinematics

We present an empirical method to measure the halo mass function (HMF) of galaxies. We determine the relation between the \hi\ line-width from single-dish observations and the dark matter halo mass ($M_{200}$) inferred from rotation curve fits in the SPARC database, then we apply this relation to galaxies from the \hi\ Parkes All Sky Survey (HIPASS) to derive the HMF. This empirical HMF is well fit by a Schecther function, and matches that expected in $Λ$CDM over the range $10^{10.5} < M_{200} < 10^{12}\;\mathrm{M}_{\odot}$. More massive halos must be poor in neutral gas to maintain consistency with the power law predicted by $Λ$CDM. We detect no discrepancy at low masses. The lowest halo mass probed by HIPASS, however, is just greater than the mass scale where the Local Group missing satellite problem sets in. The integrated mass density associated with the dark matter halos of \hi-detected galaxies sums to $Ω_{\rm m,gal} \approx 0.03$ over the probed mass range.

astro-ph.GA↗

arXiv2019

Expected Sarsa($λ$) with Control Variate for Variance Reduction

Off-policy learning is powerful for reinforcement learning. However, the high variance of off-policy evaluation is a critical challenge, which causes off-policy learning falls into an uncontrolled instability. In this paper, for reducing the variance, we introduce control variate technique to $\mathtt{Expected}$ $\mathtt{Sarsa}$($λ$) and propose a tabular $\mathtt{ES}$($λ$)-$\mathtt{CV}$ algorithm. We prove that if a proper estimator of value function reaches, the proposed $\mathtt{ES}$($λ$)-$\mathtt{CV}$ enjoys a lower variance than $\mathtt{Expected}$ $\mathtt{Sarsa}$($λ$). Furthermore, to extend $\mathtt{ES}$($λ$)-$\mathtt{CV}$ to be a convergent algorithm with linear function approximation, we propose the $\mathtt{GES}$($λ$) algorithm under the convex-concave saddle-point formulation. We prove that the convergence rate of $\mathtt{GES}$($λ$) achieves $\mathcal{O}(1/T)$, which matches or outperforms lots of state-of-art gradient-based algorithms, but we use a more relaxed condition. Numerical experiments show that the proposed algorithm performs better with lower variance than several state-of-art gradient-based TD learning algorithms: $\mathtt{GQ}$($λ$), $\mathtt{GTB}$($λ$) and $\mathtt{ABQ}$($ζ$).

cs.LG↗

arXiv2019

Gradient Q$(σ, λ)$: A Unified Algorithm with Function Approximation for Reinforcement Learning

Full-sampling (e.g., Q-learning) and pure-expectation (e.g., Expected Sarsa) algorithms are efficient and frequently used techniques in reinforcement learning. Q$(σ,λ)$ is the first approach unifies them with eligibility trace through the sampling degree $σ$. However, it is limited to the tabular case, for large-scale learning, the Q$(σ,λ)$ is too expensive to require a huge volume of tables to accurately storage value functions. To address above problem, we propose a GQ$(σ,λ)$ that extends tabular Q$(σ,λ)$ with linear function approximation. We prove the convergence of GQ$(σ,λ)$. Empirical results on some standard domains show that GQ$(σ,λ)$ with a combination of full-sampling with pure-expectation reach a better performance than full-sampling and pure-expectation methods.

cs.LG↗

arXiv2019

Dynamical Regularities in Galaxies

Galaxies are observed to obey a strict set of dynamical scaling relations. We review these relations for rotationally supported disk galaxies spanning many decades in mass, surface brightness, and gas content. The behavior of these widely varied systems can be summarized with a handful of empirical laws connected by a common acceleration scale.

astro-ph.GA↗

arXiv2019

Full-semiparametric-likelihood-based inference for non-ignorable missing data

During the past few decades, missing-data problems have been studied extensively, with a focus on the ignorable missing case, where the missing probability depends only on observable quantities. By contrast, research into non-ignorable missing data problems is quite limited. The main difficulty in solving such problems is that the missing probability and the regression likelihood function are tangled together in the likelihood presentation, and the model parameters may not be identifiable even under strong parametric model assumptions. In this paper we discuss a semiparametric model for non-ignorable missing data and propose a maximum full semiparametric likelihood estimation method, which is an efficient combination of the parametric conditional likelihood and the marginal nonparametric biased sampling likelihood. The extra marginal likelihood contribution can not only produce efficiency gain but also identify the underlying model parameters without additional assumptions. We further show that the proposed estimators for the underlying parameters and the response mean are semiparametrically efficient. Extensive simulations and a real data analysis demonstrate the advantage of the proposed method over competing methods.

stat.ME↗

arXiv2019

A Scalable Learned Index Scheme in Storage Systems

Index structures are important for efficient data access, which have been widely used to improve the performance in many in-memory systems. Due to high in-memory overheads, traditional index structures become difficult to process the explosive growth of data, let alone providing low latency and high throughput performance with limited system resources. The promising learned indexes leverage deep-learning models to complement existing index structures and obtain significant memory savings. However, the learned indexes fail to become scalable due to the heavy inter-model dependency and expensive retraining. To address these problems, we propose a scalable learned index scheme to construct different linear regression models according to the data distribution. Moreover, the used models are independent so as to reduce the complexity of retraining and become easy to partition and store the data into different pages, blocks or distributed systems. Our experimental results show that compared with state-of-the-art schemes, AIDEL improves the insertion performance by about 2$\times$ and provides comparable lookup performance, while efficiently supporting scalability.

cs.DB↗