arXiv subjects

Jian Wang

Publications and source records attributed to Jian Wang.

At least 955 records · Page 53Linked to original sources

arXiv2019

Saec: Similarity-Aware Embedding Compression in Recommendation Systems

Production recommendation systems rely on embedding methods to represent various features. An impeding challenge in practice is that the large embedding matrix incurs substantial memory footprint in serving as the number of features grows over time. We propose a similarity-aware embedding matrix compression method called Saec to address this challenge. Saec clusters similar features within a field to reduce the embedding matrix size. Saec also adopts a fast clustering optimization based on feature frequency to drastically improve clustering time. We implement and evaluate Saec on Numerous, the production distributed machine learning system in Tencent, with 10-day worth of feature data from QQ mobile browser. Testbed experiments show that Saec reduces the number of embedding vectors by two orders of magnitude, compresses the embedding size by ~27x, and delivers the same AUC and log loss performance.

cs.IR↗

arXiv2019

Valley filtering effect of phonons in graphene with a grain boundary

Due to their possibility to encode information and realize low-energy-consumption quantum devices, control and manipulation of the valley degree of freedom have been widely studied in electronic systems. In contrast, the phononic counterpart--valley phononics--has been largely unexplored, despite the importance in both fundamental science and practical applications. In this work, we demonstrate that the control of "valleys" is also applicable for phonons in graphene by using a grain boundary. In particular, perfect valley filtering effect is observed at certain energy windows for flexural modes and found to be closely related to the anisotropy of phonon valley pockets. Moreover, valley filtering may be further improved using Fano-like resonance. Our findings reveal the possibility of valley phononics, paving the road towards purposeful phonon engineering and future valley phononics.

cond-mat.mes-hall↗

arXiv2019

Adaptive Quantum Image Encryption Method Based on Wavelet Transform

An adaptive quantum image encryption method based on wavelet transform is designed. Since the characteristic of most information is centralized in the low frequency part after performing the wavelet transform, it reserves the image low frequency information only, so as to reduce the encryption workload. Then it encrypts the low frequency information by the random key stream generated by logistic map, the encryption process is realized by implementing XOR operation. In the decryption process, it carries out zero filling operations for high frequency coefficient to recover the decryption images which are equal to the plain images. At the same time, the relevant quantum logical circuit is designed. Statistical simulation and theoretical analysis demonstrate that the proposed quantum image encryption algorithm has higher security and lower computational complexity. So it is adaptive to the scenes that need to encrypt a large number of images during network transmission.

quant-ph↗

arXiv2019

Quantum coherence across bosonic superconductor-anomalous metal-insulator transitions

After decades of explorations, suffering from low critical temperature and subtle nature, whether a metallic ground state exists in a two-dimensional system beyond Anderson localization is still a mystery. Supremely, phase coherence could be the key that unlocks its intriguing nature. This work reveals how quantum phase coherence evolves across bosonic superconductor-metal-insulator transitions via magneto-conductance quantum oscillations in high-Tc superconducting films. A robust intervening anomalous metallic state characterized by both resistance and oscillation amplitude saturations in the low temperature regime is detected. By contrast, with decreasing temperature the oscillation amplitude monotonically grows on the superconducting side, but decreases at low temperatures on the insulating side. It suggests that the saturation of phase coherence plays a prominent role in the formation of this anomalous metallic state.

cond-mat.supr-con↗

arXiv2019

Multiclass Information Flow Propagation Control under Vehicle-to-Vehicle Communication Environments

Most existing models for information flow propagation in a vehicle-to-vehicle (V2V) communications environment are descriptive. They lack capabilities to control information flow, which may preclude their ability to meet application needs, including the need to propagate different information types simultaneously to different target locations within corresponding time delay bounds. This study proposes a queuing-based modeling approach to control the propagation of information flow of multiple classes. Two control parameters associated with a vehicle are leveraged to control the propagation performance of different information classes. A two-layer model is developed to characterize the information flow propagation wave (IFPW) under the designed queuing strategy. The upper layer is formulated as integro-differential equations to characterize the spatiotemporal information dissemination due to V2V communication. The lower layer characterizes the traffic flow dynamics using the Lighthill-Whitham-Richards model. The analytical solution of the asymptotic density of informed vehicles and the necessary condition for existence of the IFPW are derived for homogeneous traffic conditions. Numerical experiments provide insights on the impact of the mean communication service rate on information spread and its spatial coverage. Further, a numerical solution method is developed to solve the two-layer model, which aids in estimating the impacts of the control parameters in the queuing strategy on the IFPW speed under homogeneous and heterogeneous conditions. The proposed modeling approach enables controlling the propagation of information of different information classes to meet application needs, which can assist traffic managers to design effective and efficient traffic management and control strategies under V2V communications.

eess.SY↗

arXiv2019

Stereoscopic Dark Flash for Low-light Photography

In this work, we present a camera configuration for acquiring "stereoscopic dark flash" images: a simultaneous stereo pair in which one camera is a conventional RGB sensor, but the other camera is sensitive to near-infrared and near-ultraviolet instead of R and B. When paired with a "dark" flash (i.e., one having near-infrared and near-ultraviolet light, but no visible light) this camera allows us to capture the two images in a flash/no-flash image pair at the same time, all while not disturbing any human subjects or onlookers with a dazzling visible flash. We present a hardware prototype of this camera that approximates an idealized camera, and we present an imaging procedure that let us acquire dark flash stereo pairs that closely resemble those we would get from that idealized camera. We then present a technique for fusing these stereo pairs, first by performing registration and warping, and then by using recent advances in hyperspectral image fusion and deep learning to produce a final image. Because our camera configuration and our data acquisition process allow us to capture true low-noise long exposure RGB images alongside our dark flash stereo pairs, our learned model can be trained end-to-end to produce a fused image that retains the color and tone of a real RGB image while having the low-noise properties of a flash image.

cs.CV↗

arXiv2018

Quantized Scalar Fields in Curved Spacetime Background of Thick Brane

In this paper, we adopt the method of quantum fields in curved spacetime to quantize a free scalar matter field in the braneworld background whose warped factor is of the form that could generate Pöschl-Teller potential. Then we consider the interaction between the scalar field $ϕ(x)$ and a classical scalar source $ρ(x)$ with the form $\mathcal{H}_I=\sqrt{-g}\tilde{g}ρ(x)ϕ(x)$. The corresponding S-matrix is given and the number of particles generated by the source is obtained. Furthermore, we get the {particle numbers} in the cases that only ground state mode of the field could be detected and both the ground and first exited states could be detected, respectively. Finally, by the particle number density we define, we show how the extra dimension makes a difference specifically by the excited modes.

hep-th↗

arXiv2018

The maximum number of triangles in graphs without large linear forests

Let $G$ be a graph on $n$ vertices. A linear forest is a graph consisting of vertex-disjoint paths and isolated vertices. A maximum linear forest of $G$ is a subgraph of $G$ with maximum number of edges, which is a linear forest. We denote by $l(G)$ this maximum number. Let $t=\left\lfloor (k-1)/2\right \rfloor$. Recently, Ning and Wang \cite{boning} proved that if $l(G)=k-1$, then for any $k<n$ \[ e(G) \leq \max \left\{\binom{k}{2},\binom{t}{2}+t (n - t)+ c \right\}, \] where $c=0$ if $k$ is odd and $c=1$ otherwise, and the inequality is tight. In this paper, we prove that if $l(G)=k-1$ and $δ(G)=δ$ ($δ<\lfloor k/2 \rfloor$), then for any $k<n$ \[ e(G) \leq \max \left\{\binom{k-δ}{2}+δ(n-k+δ),\binom{t}{2}+t\left(n-t\right)+c \right\}. \] When $δ=0$, it reduces to Ning and Wang's result. Moreover, let $r_3(G)$ be the number of triangles in $G$. We prove that if $l(G)=k-1$ and $δ(G)= δ$, then for any $k<n$ \[ r_3(G)\leq \max \left\{\binom{k-δ}{3}+\binomδ{2}(n-k+δ),\binom{t}{3}+\binom{t}{2}\left(n-t\right)+d \right\}. \] where $d=0$ if $k$ is odd and $d=t$ otherwise.

math.CO↗

arXiv2018

A note on the Kenku's method

In this note, we present a complete and corrected version of Kenku's method in the proof of the nonexistence of 27-torsion on elliptic curves over quadratic fields and show why this method can not be applied to the cases of 32- and 24-torsion on elliptic curves over cubic fields.

math.NT↗

arXiv2018

On the small cyclic torsion of elliptic curves over cubic number fields

Merel's result on the strong uniform boundedness conjecture made it meaningful to classify the torsion part of the Mordell-Weil groups of all elliptic curves defined over number fields of fixed degree $d$. In this paper, we discuss the cyclic torsion subgroup of elliptic curves over cubic number fields. For $N=49,40,25$ or $22$, we show that $\mathbb{Z}/N\mathbb{Z}$ is not a subgroup of $E(K)_{tor}$ for any elliptic curve $E$ over a cubic number field $K$.

math.NT↗

arXiv2018

Time Fractional Poisson Equations: Representations and Estimates

In this paper, we study existence and uniqueness of strong as well as weak solutions for general time fractional Poisson equations. We show that there is an integral representation of the solutions of time fractional Poisson equations with zero initial values in terms of semigroup for the infinitesimal spatial generator ${\cal L}$ and the corresponding subordinator associated with the time fractional derivative. This integral representation has an integral kernel $q(t, x, y)$, which we call the fundamental solution for the time fractional Poisson equation, if the semigrou for ${\cal L}$ has an integral kernel. We further show that $q(t, x, y)$ can be expressed as a time fractional derivative of the fundamental solution for the homogenous time fractional equation under the assumption that the associated subordinator admits a conjugate subordinator. Moreover, when the Laplace exponent of the associated subordinator satisfies the weak scaling property and its distribution is self-decomposable, we establish two-sided estimates for the fundamental solution $q(t,x, y)$ through explicit estimates of transition density functions of subordinators.

math.AP↗

arXiv2018

The Turán problem for a family of tight linear forests

Let $\mathcal{F}$ be a family of $r$-graphs. The Turán number $ex_r(n;\mathcal{F})$ is defined to be the maximum number of edges in an $r$-graph of order $n$ that is $\mathcal{F}$-free. The famous Erdős Matching Conjecture shows that \[ ex_r(n,M_{k+1}^{(r)})= \max\left\{\binom{rk+r-1}{r},\binom{n}{r}-\binom{n-k}{r}\right\}, \] where $M_{k+1}^{(r)}$ represents the $r$-graph consisting of $k+1$ disjoint edges. Motivated by this conjecture, we consider the Turán problem for tight linear forests. A tight linear forest is an $r$-graph whose connected components are all tight paths or isolated vertices. Let $\mathcal{L}_{n,k}^{(r)}$ be the family of all tight linear forests of order $n$ with $k$ edges in $r$-graphs. In this paper, we prove that for sufficiently large $n$, \[ ex_r(n;\mathcal{L}_{n,k}^{(r)})=\max\left\{\binom{k}{r}, \binom{n}{r}-\binom{n-\left\lfloor (k-1)/r\right \rfloor}{r}\right\}+d, \] where $d=o(n^r)$ and if $r=3$ and $k=cn$ with $0<c<1$, if $r\geq 4$ and $k=cn$ with $0<c<1/2$. The proof is based on the weak regularity lemma for hypergraphs. We also conjecture that for arbitrary $k$ satisfying $k \equiv 1\ (mod\ r)$, the error term $d$ in the above result equals 0. We prove that the proposed conjecture implies the Erdős Matching Conjecture directly.

math.CO↗

arXiv2018

Rational function regression method for numerical analytic continuation

A simple method for numerical analytic continuation is developed. It is designed to analytically continue the imaginary time (Matsubara frequency) quantum Monte Carlo simulation results to the real time (real frequency) domain. Such a method is based on the Padé approximation. We modify it to be a linear regression problem, and then use bootstrapping statistics to get the averaged result and estimate the error. Unlike maximum entropy method, no prior information is needed. Test-cases have shown that the spectrum is recovered for inputs with relative error as high as 1%.

physics.comp-ph↗

arXiv2018

Heat kernel for non-local operators with variable order

Let $α(x)$ be a measurable function taking values in $ [α_1,α_2]$ for $0<\A_1\le \A_2<2$, and $κ(x,z)$ be a positive measurable function that is symmetric in $z$ and bounded between two positive constants. Under a uniform Hölder continuous assumptions on $α(x)$ and $x\mapsto κ(x,z)$, we obtain existence, upper and lower bounds, and regularity properties of the heat kernel associated with the following non-local operator of variable order $$ \LL f(x)=\int_{\R^d}\big(f(x+z)-f(x)-\langle\nabla f(x), z\rangle \I_{\{|z|\le 1\}}\big) \frac{κ(x,z)}{|z|^{d+α(x)}}\,dz. $$ In particular, we show that the operator $\LL$ generates a conservative Feller process on $\R^d$ having the strong Feller property, which is usually assumed a priori in the literature to study analytic properties of $\LL$ via probabilistic approaches. Our near-diagonal estimates and lower bound estimates of the heat kernel depend on the local behavior of index function $α(x)$, when $α(x)\equiv \A\in(0,2)$, our results recover some results by Chen and Kumagai (2003) and Chen and Zhang (2016).

math.PR↗

arXiv2018

A Unified Approach to Coupling SDEs driven by Lévy Noise and Some Applications

We present a general method to construct couplings of stochastic differential equations driven by Lévy noise in terms of coupling operators. This approach covers both coupling by reflection and refined basic coupling which are often discussed in the literature. As an application, we establish regularity results for the transition semigroups of the solutions to stochastic differential equations driven by additive Lévy noise.

math.PR↗

arXiv2018

Metric-Driven Learning of Correspondence Weighting for 2-D/3-D Image Registration

Registration of pre-operative 3-D volumes to intra-operative 2-D X-ray images is important in minimally invasive medical procedures. Rigid registration can be performed by estimating a global rigid motion that optimizes the alignment of local correspondences. However, inaccurate correspondences challenge the registration performance. To minimize their influence, we estimate optimal weights for correspondences using PointNet. We train the network directly with the criterion to minimize the registration error. We propose an objective function which includes point-to-plane correspondence-based motion estimation and projection error computation, thereby enabling the learning of a weighting strategy that optimally fits the underlying formulation of the registration task in an end-to-end fashion. For single-vertebra registration, we achieve an accuracy of 0.74$\pm$0.26 mm and highly improved robustness. The success rate is increased from 79.3 % to 94.3 % and the capture range from 3 mm to 13 mm.

cs.CV↗

arXiv2018

YouTube AV 50K: An Annotated Corpus for Comments in Autonomous Vehicles

With one billion monthly viewers, and millions of users discussing and sharing opinions, comments below YouTube videos are rich sources of data for opinion mining and sentiment analysis. We introduce the YouTube AV 50K dataset, a freely-available collections of more than 50,000 YouTube comments and metadata below autonomous vehicle (AV)-related videos. We describe its creation process, its content and data format, and discuss its possible usages. Especially, we do a case study of the first self-driving car fatality to evaluate the dataset, and show how we can use this dataset to better understand public attitudes toward self-driving cars and public reactions to the accident. Future developments of the dataset are also discussed.

cs.CL↗

arXiv2018

Learning Optimal Deep Projection of $^{18}$F-FDG PET Imaging for Early Differential Diagnosis of Parkinsonian Syndromes

Several diseases of parkinsonian syndromes present similar symptoms at early stage and no objective widely used diagnostic methods have been approved until now. Positron emission tomography (PET) with $^{18}$F-FDG was shown to be able to assess early neuronal dysfunction of synucleinopathies and tauopathies. Tensor factorization (TF) based approaches have been applied to identify characteristic metabolic patterns for differential diagnosis. However, these conventional dimension-reduction strategies assume linear or multi-linear relationships inside data, and are therefore insufficient to distinguish nonlinear metabolic differences between various parkinsonian syndromes. In this paper, we propose a Deep Projection Neural Network (DPNN) to identify characteristic metabolic pattern for early differential diagnosis of parkinsonian syndromes. We draw our inspiration from the existing TF methods. The network consists of a (i) compression part: which uses a deep network to learn optimal 2D projections of 3D scans, and a (ii) classification part: which maps the 2D projections to labels. The compression part can be pre-trained using surplus unlabelled datasets. Also, as the classification part operates on these 2D projections, it can be trained end-to-end effectively with limited labelled data, in contrast to 3D approaches. We show that DPNN is more effective in comparison to existing state-of-the-art and plausible baselines.

cs.CV↗