arXiv subjects

Hong Chang

Publications and source records attributed to Hong Chang.

At least 91 records · Page 5Linked to original sources

arXiv2019

Scalable heterostructures produced through mechanical abrasion of van der Waals powders

To fully exploit van der Waals materials and heterostructures, new mass-scalable production routes that are low cost but preserve the high electronic and optical quality of the single crystals are required. Here, we demonstrate an approach to realize a variety of functional heterostructures based on van der Waals nanocrystal films produced through the mechanical abrasion of bulk powders. Significant performance improvements are realized in our devices compared to those fabricated through ink-jet printing of nanocrystal dispersions. To highlight the simplicity and scalability of the technology a multitude of different functional heterostructure devices such as resistors, capacitors, photovoltaics as well as energy devices such as large-area catalyst coatings for hydrogen evolution reaction and multilayer heterostructures for triboelectric nanogenerators are shown. The simplicity of the device fabrication, scalability, and compatibility with flexible substrates makes this a promising technological route for up-scalable van der Waals heterostructures.

physics.app-ph↗

arXiv2019

Cross Attention Network for Few-shot Classification

Few-shot classification aims to recognize unlabeled samples from unseen classes given only few labeled samples. The unseen classes and low-data problem make few-shot classification very challenging. Many existing approaches extracted features from labeled and unlabeled samples independently, as a result, the features are not discriminative enough. In this work, we propose a novel Cross Attention Network to address the challenging problems in few-shot classification. Firstly, Cross Attention Module is introduced to deal with the problem of unseen classes. The module generates cross attention maps for each pair of class feature and query sample feature so as to highlight the target object regions, making the extracted feature more discriminative. Secondly, a transductive inference algorithm is proposed to alleviate the low-data problem, which iteratively utilizes the unlabeled query set to augment the support set, thereby making the class features more representative. Extensive experiments on two benchmarks show our method is a simple, effective and computationally efficient framework and outperforms the state-of-the-arts.

cs.CV↗

arXiv2019

Temporal Knowledge Propagation for Image-to-Video Person Re-identification

In many scenarios of Person Re-identification (Re-ID), the gallery set consists of lots of surveillance videos and the query is just an image, thus Re-ID has to be conducted between image and videos. Compared with videos, still person images lack temporal information. Besides, the information asymmetry between image and video features increases the difficulty in matching images and videos. To solve this problem, we propose a novel Temporal Knowledge Propagation (TKP) method which propagates the temporal knowledge learned by the video representation network to the image representation network. Specifically, given the input videos, we enforce the image representation network to fit the outputs of video representation network in a shared feature space. With back propagation, temporal knowledge can be transferred to enhance the image features and the information asymmetry problem can be alleviated. With additional classification and integrated triplet losses, our model can learn expressive and discriminative image and video features for image-to-video re-identification. Extensive experiments demonstrate the effectiveness of our method and the overall results on two widely used datasets surpass the state-of-the-art methods by a large margin. Code is available at: https://github.com/guxinqian/TKP

cs.CV↗

arXiv2019

VRSTC: Occlusion-Free Video Person Re-Identification

Video person re-identification (re-ID) plays an important role in surveillance video analysis. However, the performance of video re-ID degenerates severely under partial occlusion. In this paper, we propose a novel network, called Spatio-Temporal Completion network (STCnet), to explicitly handle partial occlusion problem. Different from most previous works that discard the occluded frames, STCnet can recover the appearance of the occluded parts. For one thing, the spatial structure of a pedestrian frame can be used to predict the occluded body parts from the unoccluded body parts of this frame. For another, the temporal patterns of pedestrian sequence provide important clues to generate the contents of occluded parts. With the Spatio-temporal information, STCnet can recover the appearance for the occluded parts, which could be leveraged with those unoccluded parts for more accurate video re-ID. By combining a re-ID network with STCnet, a video re-ID framework robust to partial occlusion (VRSTC) is proposed. Experiments on three challenging video re-ID databases demonstrate that the proposed approach outperforms the state-of-the-art.

cs.CV↗

arXiv2019

Interaction-and-Aggregation Network for Person Re-identification

Person re-identification (reID) benefits greatly from deep convolutional neural networks (CNNs) which learn robust feature embeddings. However, CNNs are inherently limited in modeling the large variations in person pose and scale due to their fixed geometric structures. In this paper, we propose a novel network structure, Interaction-and-Aggregation (IA), to enhance the feature representation capability of CNNs. Firstly, Spatial IA (SIA) module is introduced. It models the interdependencies between spatial features and then aggregates the correlated features corresponding to the same body parts. Unlike CNNs which extract features from fixed rectangle regions, SIA can adaptively determine the receptive fields according to the input person pose and scale. Secondly, we introduce Channel IA (CIA) module which selectively aggregates channel features to enhance the feature representation, especially for smallscale visual cues. Further, IA network can be constructed by inserting IA blocks into CNNs at any depth. We validate the effectiveness of our model for person reID by demonstrating its superiority over state-of-the-art methods on three benchmark datasets.

cs.CV↗

arXiv2019

Cascade RetinaNet: Maintaining Consistency for Single-Stage Object Detection

Recent researches attempt to improve the detection performance by adopting the idea of cascade for single-stage detectors. In this paper, we analyze and discover that inconsistency is the major factor limiting the performance. The refined anchors are associated with the feature extracted from the previous location and the classifier is confused by misaligned classification and localization. Further, we point out two main designing rules for the cascade manner: improving consistency between classification confidence and localization performance, and maintaining feature consistency between different stages. A multistage object detector named Cas-RetinaNet, is then proposed for reducing the misalignments. It consists of sequential stages trained with increasing IoU thresholds for improving the correlation, and a novel Feature Consistency Module for mitigating the feature inconsistency. Experiments show that our proposed Cas-RetinaNet achieves stable performance gains across different models and input scales. Specifically, our method improves RetinaNet from 39.1 AP to 41.1 AP on the challenging MS COCO dataset without any bells or whistles.

cs.CV↗

arXiv2019

Reevaluation of the nuclear electric quadrupole moment for 87Sr by hyperfine structures and relativistic atomic theory

The values of nuclear electric quadrupole moment are different by about 7% for 87Sr nucleus between the recommended value [N. J. Stone, At. Data Nucl. Data Tables 111-112, 1 (2016); P. Pyykko, Mol. Phys. 116, 1328 (2018)] and earlier results [e.g. A. M. Matensson-Pendrill, J. Phys. B: At. Mol. Opt. Phys. 35, 917 (2002); K. Z. Yu et al., Phys. Rev. A 70, 012506 (2004)]. In this work, we reported a new value, Q(87Sr) = 328(4) mb, making use of our calculated electric field gradients produced by electrons at nucleus in combination with experimental values for hyperfine structures of the 5s5p 3P1,2 states of the neutral Sr atom. In the framework of the multi-configuration Dirac-Hartree-Fock theory, the electron correlations were taken into account systematically so as to control the uncertainties of the electric field gradient at about 1% level. The present result is different from the recommended value, but in excellent agreement with those by Matensson-Pendrill and Yu et al.. We would recommend the present Q value as a reference for 87Sr.

physics.atom-ph↗

arXiv2019

WIDER Face and Pedestrian Challenge 2018: Methods and Results

This paper presents a review of the 2018 WIDER Challenge on Face and Pedestrian. The challenge focuses on the problem of precise localization of human faces and bodies, and accurate association of identities. It comprises of three tracks: (i) WIDER Face which aims at soliciting new approaches to advance the state-of-the-art in face detection, (ii) WIDER Pedestrian which aims to find effective and efficient approaches to address the problem of pedestrian detection in unconstrained environments, and (iii) WIDER Person Search which presents an exciting challenge of searching persons across 192 movies. In total, 73 teams made valid submissions to the challenge tracks. We summarize the winning solutions for all three tracks. and present discussions on open problems and potential research directions in these topics.

cs.CV↗

arXiv2018

More on limited packings in graphs

A set $B$ of vertices in a graph $G$ is called a \emph{$k$-limited packing} if for each vertex $v$ of $G$, its closed neighbourhood has at most $k$ vertices in $B$. The \emph{$k$-limited packing number} of a graph $G$, denoted by $L_k(G)$, is the largest number of vertices in a $k$-limited packing in $G$. The concept of the $k$-limited packing of a graph was introduced by Gallant et al., which is a generalization of the well-known packing of a graph. In this paper, we present some tight bounds for the $k$-limited packing number of a graph in terms of its order, diameter, girth, and maximum degree, respectively. As a result, we obtain the tight Nordhaus-Gaddum-type result of this parameter for general $k$. At last, we investigate the relationship among the open packing number, the packing number and $2$-limited packing number of trees.

math.CO↗

arXiv2018

Conflict-free connection of trees

We study the conflict-free connection coloring of trees, which is also the conflict-free coloring of the so-called edge-path hypergraphs of trees. We first prove that for a tree $T$ of order $n$, $cfc(T)\geq cfc(P_n)=\lceil \log_{2} n\rceil$, which completely confirms the conjecture of Li and Wu. We then present a sharp upper bound for the conflict-free connection number of trees by a simple algorithm. Furthermore, we show that the conflict-free connection number of the binomial tree with $2^{k-1}$ vertices is $k-1$. At last, we study trees which are $cfc$-critical, and prove that if a tree $T$ is $cfc$-critical, then the conflict-free connection coloring of $T$ is equivalent to the edge ranking of $T$.

math.CO↗

arXiv2018

Graphs with conflict-free connection number two

An edge-colored graph $G$ is \emph{conflict-free connected} if any two of its vertices are connected by a path, which contains a color used on exactly one of its edges. The \emph{conflict-free connection number} of a connected graph $G$, denoted by $cfc(G)$, is the smallest number of colors needed in order to make $G$ conflict-free connected. For a graph $G,$ let $C(G)$ be the subgraph of $G$ induced by its set of cut-edges. In this paper, we first show that, if $G$ is a connected non-complete graph $G$ of order $n\geq 9$ with $C(G)$ being a linear forest and with the minimum degree %$δ(G)\geq 2$, then $cfc(G)=2$ for $4 \leq n\leq 8 $; if $δ(G)\geq \max\{3, \frac{n-4}{5}\}$, then $cfc(G)=2$. The bound on the minimum degree is best possible. Next, we prove that, if $G$ is a connected non-complete graph of order $n\geq 33$ with $C(G)$ being a linear forest and with $d(x)+d(y)\geq \frac{2n-9}{5}$ for each pair of two nonadjacent vertices $x, y$ of $V(G)$, then $cfc(G)=2$. Both bounds, on the order $n$ and the degree sum, are tight. Moreover, we prove several results concerning relations between degree conditions on $G$ and the number of cut edges in $G$.

math.CO↗

arXiv2018

High-order sideband effect with two-level atoms in a shaken optical lattice

The system of cold two-level atoms inside an optical lattice potential has been used to simulate various models encountered in fundamental and condensed-matter physics. When the optical lattice is periodically shaken, some interesting dynamical effects were observed. In this work, by utilizing the ultra-narrow linewidth of the dipole-forbidden transition in $^{87}$Sr atoms loaded in a shaken optical lattice, we reach a new operation regime where the shaking rate is much larger than linewidth of the two-level transition but much smaller than the frequency interval of the harmonic lattice potential, which was not realizable before. Under such unique conditions, an interesting quantum dynamical effect, the high-order sideband effect (HSE), has been experimentally observed. All the results can be well explained by a developed theoretical model with analytical solution. This HSE is governed by different mechanism as the similar phenomenon observed in semiconductors and has unique characteristics.

physics.atom-ph↗

arXiv2017

Exploration of the magnetic-field-induced $5s5p$~$^3P_0$~-~$5s^2$~$^1S_0$ forbidden transition in bosonic Sr atom

We presented experimental and theoretical studies of the effect of an external magnetic field on the forbidden transition in the bosonic Sr atom. In our ultra-cold atomic system, the excitation fraction of $5s5p$~$^3P_0$~-~$5s^2$~$^1S_0$ forbidden transition was measured under the circumstance of the different magnetic field strengths by using the normalized detection method. Based on perturbation theory, we calculated the magnetic-field-induced $5s5p$~$^3P_0$~-~$5s^2$~$^1S_0$ transition rate. The excitation fraction as a function of the magnetic field strength was deduced according to the calculated results. A good agreement was found between the experimental measurements and the calculations. This study should be helpful in evaluating the magnetic field effects on the forbidden transition rate with higher accuracy. Moreover, it can help to understand the ultra-cold atomic interaction in the external magnetic field.

physics.atom-ph↗

arXiv2017

Super-Resolution with Deep Adaptive Image Resampling

Deep learning based methods have recently pushed the state-of-the-art on the problem of Single Image Super-Resolution (SISR). In this work, we revisit the more traditional interpolation-based methods, that were popular before, now with the help of deep learning. In particular, we propose to use a Convolutional Neural Network (CNN) to estimate spatially variant interpolation kernels and apply the estimated kernels adaptively to each position in the image. The whole model is trained in an end-to-end manner. We explore two ways to improve the results for the case of large upscaling factors, and propose a recursive extension of our basic model. This achieves results that are on par with state-of-the-art methods. We visualize the estimated adaptive interpolation kernels to gain more insight on the effectiveness of the proposed method. We also extend the method to the task of joint image filtering and again achieve state-of-the-art performance.

cs.CV↗

arXiv2017

Nordhaus-Gaddum-type theorem for conflict-free connection number of graphs

An edge-colored graph $G$ is \emph{conflict-free connected} if, between each pair of distinct vertices, there exists a path containing a color used on exactly one of its edges. The \emph{conflict-free connection number} of a connected graph $G$, denoted by $cfc(G)$, is defined as the smallest number of colors that are needed in order to make $G$ conflict-free connected. In this paper, we determine all trees $T$ of order $n$ for which $cfc(T)=n-t$, where $t\geq 1$ and $n\geq 2t+2 $. Then we prove that $1\leq cfc(G)\leq n-1$ for a connected graph $G$, and characterize the graphs $G$ with $cfc(G)=1,n-4,n-3,n-2,n-1$, respectively. Finally, we get the Nordhaus-Gaddum-type theorem for the conflict-free connection number of graphs, and prove that if $G$ and $\overline{G}$ are connected, then $4\leq cfc(G)+cfc(\overline{G})\leq n$ and $4\leq cfc(G)\cdot cfc(\overline{G})\leq2(n-2)$, and moreover, $cfc(G)+cfc(\overline{G})=n$ or $cfc(G)\cdot cfc(\overline{G})=2(n-2)$ if and only if one of $G$ and $\overline{G}$ is a tree with maximum degree $n-2$ or a $P_5$, and the lower bounds are sharp.

math.CO↗

arXiv2017

More on the $k$-color connection number of a graph

An edge-colored graph $G$ is $k$-color connected if, between each pair of vertices, there exists a path using at least $k$ different colors. The $k$-color connection number of $G$, denoted by $cc_{k}(G)$, is the minimum number of colors needed to color the edges of $G$ so that $G$ is $k$-color connected. First, we prove that let $H$ be a subdivision of a connected graph $G$, then $cc_{k}(H)\leq cc_{k}(G)$. Second, we give sufficient conditions to guarantee that $cc_{k}(G)=k$ in terms of minimum degree and the number of edges for 2-connected graphs. As a byproduct, we show that almost all graphs have the $k$-color connection number $k$. At last, we investigate the relationship between the $k$-color connection number and the rainbow connection number for a connected graph. In addition, we give exact values of $k$-color connection numbers for some graph classes: subdivisions of the wheel and the complete graph, and the generalised $θ$-graph.

math.CO↗

arXiv2016

Degree sum conditions for graphs to have proper connection number 2

A path $P$ in an edge-colored graph $G$ is a \emph{proper path} if no two adjacent edges of $P$ are colored with the same color. The graph $G$ is \emph{proper connected} if, between every pair of vertices, there exists a proper path in $G$. The \emph{proper connection number} $pc(G)$ of a connected graph $G$ is defined as the minimum number of colors to make $G$ proper connected. In this paper, we study the degree sum condition for a general graph or a bipartite graph to have proper connection number 2. First, we show that if $G$ is a connected noncomplete graph of order $n\geq 5$ such that $d(x)+d(y)\geq \frac{n}{2}$ for every pair of nonadjacent vertices $x,y\in V(G)$, then $pc(G)=2$ except for three small graphs on 6, 7 and 8 vertices. In addition, we obtain that if $G$ is a connected bipartite graph of order $n\geq 4$ such that $d(x)+d(y)\geq \frac{n+6}{4}$ for every pair of nonadjacent vertices $x,y\in V(G)$, then $pc(G)=2$. Examples are given to show that the above conditions are best possible.

math.CO↗

arXiv2016

The $(k,\ell)$-proper index of graphs

A tree $T$ in an edge-colored graph is called a {\it proper tree} if no two adjacent edges of $T$ receive the same color. Let $G$ be a connected graph of order $n$ and $k$ be an integer with $2\leq k \leq n$. For $S\subseteq V(G)$ and $|S| \ge 2$, an $S$-tree is a tree containing the vertices of $S$ in $G$. Suppose $\{T_1,T_2,\ldots,T_\ell\}$ is a set of $S$-trees, they are called \emph{internally disjoint} if $E(T_i)\cap E(T_j)=\emptyset$ and $V(T_i)\cap V(T_j)=S$ for $1\leq i\neq j\leq \ell$. For a set $S$ of $k$ vertices of $G$, the maximum number of internally disjoint $S$-trees in $G$ is denoted by $κ(S)$. The $κ$-connectivity $κ_k(G)$ of $G$ is defined by $κ_k(G)=\min\{κ(S)\mid S$ is a $k$-subset of $V(G)\}$. For a connected graph $G$ of order $n$ and for two integers $k$ and $\ell$ with $2\le k\le n$ and $1\leq \ell \leq κ_k(G)$, the \emph{$(k,\ell)$-proper index $px_{k,\ell}(G)$} of $G$ is the minimum number of colors that are needed in an edge-coloring of $G$ such that for every $k$-subset $S$ of $V(G)$, there exist $\ell$ internally disjoint proper $S$-trees connecting them. In this paper, we show that for every pair of positive integers $k$ and $\ell$ with $k \ge 3$, there exists a positive integer $N_1=N_1(k,\ell)$ such that $px_{k,\ell}(K_n) = 2$ for every integer $n \ge N_1$, and also there exists a positive integer $N_2=N_2(k,\ell)$ such that $px_{k,\ell}(K_{m,n}) = 2$ for every integer $n \ge N_2$ and $m=O(n^r) (r \ge 1)$. In addition, we show that for every $p \ge c\sqrt[k]{\frac{\log_a n}{n}}$ ($c \ge 5$), $px_{k,\ell}(G_{n,p})\le 2$ holds almost surely, where $G_{n,p}$ is the Erdös-Rényi random graph model.

math.CO↗