arXiv subjects

Chang Xu

Publications and source records attributed to Chang Xu.

At least 415 records · Page 23Linked to original sources

arXiv2017

Privileged Multi-label Learning

This paper presents privileged multi-label learning (PrML) to explore and exploit the relationship between labels in multi-label learning problems. We suggest that for each individual label, it cannot only be implicitly connected with other labels via the low-rank constraint over label predictors, but also its performance on examples can receive the explicit comments from other labels together acting as an \emph{Oracle teacher}. We generate privileged label feature for each example and its individual label, and then integrate it into the framework of low-rank based multi-label learning. The proposed algorithm can therefore comprehensively explore and exploit label relationships by inheriting all the merits of privileged information and low-rank constraints. We show that PrML can be efficiently solved by dual coordinate descent algorithm using iterative optimization strategy with cheap updates. Experiments on benchmark datasets show that through privileged label features, the performance can be significantly improved and PrML is superior to several competing methods in most cases.

stat.ML↗

arXiv2016

DroidLeaks: Benchmarking Resource Leak Bugs for Android Applications

Resource leak bugs in Android apps are pervasive and can cause serious performance degradation and system crashes. In recent years, several resource leak detection techniques have been proposed to assist Android developers in correctly managing system resources. Yet, there exist no common bug benchmarks for effectively and reliably comparing such techniques and quantitatively evaluating their strengths and weaknesses. This paper describes our initial contribution towards constructing such a benchmark. To locate real resource leak bugs, we mined 124,215 code revisions of 34 large-scale open-source Android apps. We successfully found 298 fixed resource leaks, which cover a diverse set of resource classes, from 32 out of the 34 apps. To understand the characteristics of these bugs, we conducted an empirical study, which revealed the root causes of frequent resource leaks in Android apps and common patterns of faults made by developers. With our findings, we further implemented a static checker to detect a common pattern of resource leaks in Android apps. Experiments showed that the checker can effectively locate real resource leaks in popular Android apps, confirming the usefulness of our work.

cs.SE↗

arXiv2016

A new approach for calculating nuclear symmetry energy

By using the functional path integral method, we obtain a model-independent formula for nuclear symmetry energy, which explicitly shows the relation between nuclear symmetry energy and isospin susceptibility. The latter one is found to be a probe to the QCD chiral phase transition. We further found that, the nuclear symmetry energy has an abrupt change at the critical nuclear density where the chiral symmetry restores partially, which could be detected from the experiments.

nucl-th↗

arXiv2016

Streaming View Learning

An underlying assumption in conventional multi-view learning algorithms is that all views can be simultaneously accessed. However, due to various factors when collecting and pre-processing data from different views, the streaming view setting, in which views arrive in a streaming manner, is becoming more common. By assuming that the subspaces of a multi-view model trained over past views are stable, here we fine tune their combination weights such that the well-trained multi-view model is compatible with new views. This largely overcomes the burden of learning new view functions and updating past view functions. We theoretically examine convergence issues and the influence of streaming views in the proposed algorithm. Experimental results on real-world datasets suggest that studying the streaming views problem in multi-view learning is significant and that the proposed algorithm can effectively handle streaming views in different applications.

stat.ML↗

arXiv2016

Streaming Label Learning for Modeling Labels on the Fly

It is challenging to handle a large volume of labels in multi-label learning. However, existing approaches explicitly or implicitly assume that all the labels in the learning process are given, which could be easily violated in changing environments. In this paper, we define and study streaming label learning (SLL), i.e., labels are arrived on the fly, to model newly arrived labels with the help of the knowledge learned from past labels. The core of SLL is to explore and exploit the relationships between new labels and past labels and then inherit the relationship into hypotheses of labels to boost the performance of new classifiers. In specific, we use the label self-representation to model the label relationship, and SLL will be divided into two steps: a regression problem and a empirical risk minimization (ERM) problem. Both problems are simple and can be efficiently solved. We further show that SLL can generate a tighter generalization error bound for new labels than the general ERM framework with trace norm or Frobenius norm regularization. Finally, we implement extensive experiments on various benchmark datasets to validate the new setting. And results show that SLL can effectively handle the constantly emerging new labels and provides excellent classification performance.

stat.ML↗

arXiv2016

Parts for the Whole: The DCT Norm for Extreme Visual Recovery

Here we study the extreme visual recovery problem, in which over 90\% of pixel values in a given image are missing. Existing low rank-based algorithms are only effective for recovering data with at most 90\% missing values. Thus, we exploit visual data's smoothness property to help solve this challenging extreme visual recovery problem. Based on the Discrete Cosine Transformation (DCT), we propose a novel DCT norm that involves all pixels and produces smooth estimations in any view. Our theoretical analysis shows that the total variation (TV) norm, which only achieves local smoothness, is a special case of the proposed DCT norm. We also develop a new visual recovery algorithm by minimizing the DCT and nuclear norms to achieve a more visually pleasing estimation. Experimental results on a benchmark image dataset demonstrate that the proposed approach is superior to state-of-the-art methods in terms of peak signal-to-noise ratio and structural similarity.

cs.CV↗

arXiv2016

Theoretical study on nuclear structure by the multiple Coulomb scattering and magnetic scattering of relativistic electrons

Electron scattering is an effective method to study the nuclear structure. For the odd-$A$ nuclei with proton holes in the outmost orbits, we investigate the contributions of proton holes to the nuclear quadrupole moments $Q$ and magnetic moments $μ$ by the multiple Coulomb scattering and magnetic scattering. The deformed nuclear charge densities are constructed by the relativistic mean-field (RMF) models. Comparing the theoretical Coulomb and magnetic form factors with the experimental data, the nuclear quadrupole moments $Q$ and nuclear magnetic moments $μ$ are investigated. From the electron scattering, the wave functions of the proton holes of odd-$A$ nuclei can be tested, which can also reflect the validity of the nuclear structure model.

nucl-th↗

arXiv2015

Investigation of 10Be and its cluster dynamics from nonlocalized clustering concept

We extend the new concept of nonlocalized clustering to the nucleus 10Be with proton number Z=4 and neutron number N=6 (N=Z+2). The Tohsaki-Horiuchi-Schuck-Röpke (THSR) wave function is formulated for the description of different structures of 10Be. Physical properties such as energy spectrum and root-mean-square radii are calculated for the first two 0+ states and corresponding rotational bands. With only one single THSR wave function, the calculated results show good agreement with other models and experimental values. We apply, for the first time, the THSR wave function on the chain orbit (σ-orbit) structure in the 0^+_2 state of 10Be. The ring orbit (π-orbit) and σ-orbit structures are further illustrated by calculating the density distribution of the valence neutrons. We also investigate the dynamics of ff-clusters and the correlations of two valence neutrons in 10Be.

nucl-th↗

arXiv2015

Alpha Decay Width of $^{212}$Po from a quartetting wave function approach

A microscopic calculation of $α$-cluster preformation probability and $α$ decay width in the typical $α$ emitter $^{212}$Po is presented. Results are obtained by improving a recent approach to describe $α$ preformation in $^{212}$Po [Phys. Rev. C 90, 034304 (2014)] implementing four-nucleon correlations (quartetting). Using the actually measured density distribution of the $^{208}$ Pb core, the calculated alpha decay width of $^{212}$Po agrees fairly well with the measured one.

nucl-th↗

arXiv2014

Investigation of ${}^9$Be from nonlocalized clustering concept

The nonlocalized aspect of clustering, which is a new concept for self-conjugate nuclei, is extended for the investigation of the N{\not=}Z nucleus ${}^9$Be. A modified version of the THSR (Tohsaki-Horiuchi-Schuck-Röpke) wave function is introduced based on the container picture. It is found that the constructed negative-parity THSR wave function is very suitable for describing the cluster states of ${}^9$Be. Namely the nonlocalized clustering is shown to prevail in ${}^9$Be. The calculated binding energy and radius of ${}^9$Be are consistent with calculations in other models and with experimental values. The squared overlaps between the THSR wave function and the Brink+GCM wave function for the $3/2^-$ rotational band of ${}^9$Be are found to be near 96%. Furthermore, by showing the density distribution of the ground state of ${}^9$Be, the π-orbit structure is naturally reproduced by using this THSR wave function.

nucl-th↗

arXiv2014

Local Rademacher Complexity for Multi-label Learning

We analyze the local Rademacher complexity of empirical risk minimization (ERM)-based multi-label learning algorithms, and in doing so propose a new algorithm for multi-label learning. Rather than using the trace norm to regularize the multi-label predictor, we instead minimize the tail sum of the singular values of the predictor in multi-label learning. Benefiting from the use of the local Rademacher complexity, our algorithm, therefore, has a sharper generalization error bound and a faster convergence rate. Compared to methods that minimize over all singular values, concentrating on the tail singular values results in better recovery of the low-rank structure of the multi-label predictor, which plays an import role in exploiting label correlations. We propose a new conditional singular value thresholding algorithm to solve the resulting objective function. Empirical studies on real-world datasets validate our theoretical results and demonstrate the effectiveness of the proposed algorithm.

stat.ML↗

arXiv2014

Convex hulls of random walks and their scaling limits

For the perimeter length and the area of the convex hull of the first $n$ steps of a planar random walk, we study $n \to \infty$ mean and variance asymptotics and establish non-Gaussian distributional limits. Our results apply to random walks with drift (for the area) and walks with no drift (for both area and perimeter length) under mild moments assumptions on the increments. These results complement and contrast with previous work which showed that the perimeter length in the case with drift satisfies a central limit theorem. We deduce these results from weak convergence statements for the convex hulls of random walks to scaling limits defined in terms of convex hulls of certain Brownian motions. We give bounds that confirm that the limiting variances in our results are non-zero.

math.PR↗

arXiv2014

Sequential Click Prediction for Sponsored Search with Recurrent Neural Networks

Click prediction is one of the fundamental problems in sponsored search. Most of existing studies took advantage of machine learning approaches to predict ad click for each event of ad view independently. However, as observed in the real-world sponsored search system, user's behaviors on ads yield high dependency on how the user behaved along with the past time, especially in terms of what queries she submitted, what ads she clicked or ignored, and how long she spent on the landing pages of clicked ads, etc. Inspired by these observations, we introduce a novel framework based on Recurrent Neural Networks (RNN). Compared to traditional methods, this framework directly models the dependency on user's sequential behaviors into the click prediction process through the recurrent structure in RNN. Large scale evaluations on the click-through logs from a commercial search engine demonstrate that our approach can significantly improve the click prediction accuracy, compared to sequence-independent approaches.

cs.IR↗

arXiv2014

Bound clusters on top of doubly magic nuclei

An effective $α$ particle equation is derived for cases where an $α$ particle is formed on top of a doubly magic nucleus. As an example, we consider $^{212}$Po with the $α$ on top of the $^{208}$ Pb core. We will consider the core nucleus infinitely heavy, so that the $α$ particle moves with respect to a fixed center, i.e., recoil effects are neglected. The fully quantal solution of the problem is discussed. The approach is inspired by the THSR (Tohsaki-Horiuchi-Schuck-Röpke) wave function concept that has been successfully applied to light nuclei. Shell model calculations are improved by including four-particle ($α$-like) correlations that are of relevance when the matter density becomes low. In the region where the $α$-like cluster penetrates the core nucleus, the intrinsic bound state wave function transforms at a critical density into an unbound four-nucleon shell model state. Exploratory calculations for $^{212}$Po are presented. Such preformed cluster states are only hardly described by shell model calculations. Reasons for different physics behavior of an $α$-like cluster with respect to a deuteron-like cluster are discussed.

nucl-th↗

arXiv2014

Probing isospin- and momentum-dependent nuclear effective interactions in neutron-rich matter

The single-particle potentials for nucleons and hyperons in neutron-rich matter generally depends on the density and isospin asymmetry of the medium as well as the momentum and isospin of the particle. It further depends on the temperature of the matter if the latter is in thermal equilibrium. We review here the extension of a Gogny-type isospin- and momentum-dependent interaction in several aspects made in recent years and their applications in studying intermediate-energy heavy ion collisions, thermal properties of asymmetric nuclear matter and properties of neutron stars. The importance of the isospin- and momentum-dependence of the single-particle potential, especially the momentum dependence of the isovector potential, is clearly revealed throughout these studies.

nucl-th↗

arXiv2013

Nonlocalized cluster dynamics and nuclear molecular structure

A container picture is proposed for understanding cluster dynamics where the clusters make nonlocalized motion occupying the lowest orbit of the cluster mean-field potential characterized by the size parameter $``B"$ in the THSR (Tohsaki-Horiuchi-Schuck-Röpke) wave function. The nonlocalized cluster aspects of the inversion-doublet bands in $^{20}$Ne which have been considered as a typical manifestation of localized clustering are discussed. So far unexplained puzzling features of the THSR wave function, namely that after angular-momentum projection for two cluster systems the prolate THSR wave function is almost 100$\%$ equivalent to an oblate THSR wave function is clarified. It is shown that the true intrinsic two-cluster THSR configuration is nonetheless prolate. The proposal of the container picture is based on the fact that typical cluster systems, 2$α$, 3$α$, and $α$+$^{16}$O, are all well described by a single THSR wave function. It will be shown for the case of linear-chain states with two and three $α$-clusters as well as for the $α$+$^{16}$O system that localization is entirely of kinematical origin, that is, due to the inter-cluster Pauli repulsion. It is concluded that this feature is general for nuclear cluster states.

nucl-th↗

arXiv2013

Relationship between the symmetry energy and the single-nucleon potential in isospin-asymmetric nucleonic matter

In this contribution, we review the most important physics presented originally in our recent publications. Some new analyses, insights and perspectives are also provided. We showed recently that the symmetry energy $E_{sym}(ρ)$ and its density slope $L(ρ)$ at an arbitrary density $ρ$ can be expressed analytically in terms of the magnitude and momentum dependence of the single-nucleon potentials using the Hugenholtz-Van Hove (HVH) theorem. These relationships provide new insights about the fundamental physics governing the density dependence of nuclear symmetry energy. Using the isospin and momentum (k) dependent MDI interaction as an example, the contribution of different terms in the single-nucleon potential to the $E_{sym}(ρ)$ and $L(ρ)$ are analyzed in detail at different densities. It is shown that the behavior of $E_{sym}(ρ)$ is mainly determined by the first-order symmetry potential $U_{sym,1}(ρ,k)$ of the single-nucleon potential. The density slope $L(ρ)$ depends not only on the first-order symmetry potential $U_{sym,1}(ρ,k)$ but also the second-order one $U_{sym,2}(ρ,k)$. Both the $U_{sym,1}(ρ,k)$ and $U_{sym,2}(ρ,k)$ at normal density $ρ_0$ are constrained by the isospin and momentum dependent nucleon optical potential extracted from the available nucleon-nucleus scattering data. The $U_{sym,2}(ρ,k)$ especially at high density and momentum affects significantly the $L(ρ)$, but it is theoretically poorly understood and currently there is almost no experimental constraints known.

nucl-th↗

arXiv2013

Nonlocalized Clustering: A New Concept in Nuclear Cluster Structure Physics

We investigate the $α$+\oo\ cluster structure in the inversion-doublet band ($K^π=0_{1}^\pm$) states of \nene\ with an angular-momentum-projected version of the Tohsaki-Horiuchi-Schuck-Röpke (THSR) wave function, which was successful "in its original form" for the description of, e.g., the famous Hoyle state. In contrast with the traditional view on clusters as localized objects, especially in inversion doublets, we find that these {\it single} THSR wave functions, which are based on the concept of nonlocalized clustering, can well describe the $K^π=0_1^-$ band and the $K^π=0_1^+$ band. For instance, they have 99.98% and 99.87% squared overlaps for $1^-$ and $3^- $ states (99.29%, 98.79% and 97.75% for $0^+, 2^+$ and $4^+$ states), respectively, with the corresponding exact solution of the $α$+\oo\ resonating group method. These astounding results shed a completely new light on the physics of low energy nuclear cluster states in nuclei: The clusters are nonlocalized and move around in the whole nuclear volume, only avoiding mutual overlap due to the Pauli blocking effect.

nucl-th↗