arXiv subjects

Huan Li

Publications and source records attributed to Huan Li.

At least 145 records · Page 8Linked to original sources

arXiv2019

On the Complexity Analysis of the Primal Solutions for the Accelerated Randomized Dual Coordinate Ascent

Dual first-order methods are essential techniques for large-scale constrained convex optimization. However, when recovering the primal solutions, we need $T(ε^{-2})$ iterations to achieve an $ε$-optimal primal solution when we apply an algorithm to the non-strongly convex dual problem with $T(ε^{-1})$ iterations to achieve an $ε$-optimal dual solution, where $T(x)$ can be $x$ or $\sqrt{x}$. In this paper, we prove that the iteration complexity of the primal solutions and dual solutions have the same $O\left(\frac{1}{\sqrtε}\right)$ order of magnitude for the accelerated randomized dual coordinate ascent. When the dual function further satisfies the quadratic functional growth condition, by restarting the algorithm at any period, we establish the linear iteration complexity for both the primal solutions and dual solutions even if the condition number is unknown. When applied to the regularized empirical risk minimization problem, we prove the iteration complexity of $O\left(n\log n+\sqrt{\frac{n}ε}\right)$ in both primal space and dual space, where $n$ is the number of samples. Our result takes out the $\left(\log \frac{1}ε\right)$ factor compared with the methods based on smoothing/regularization or Catalyst reduction. As far as we know, this is the first time that the optimal $O\left(\sqrt{\frac{n}ε}\right)$ iteration complexity in the primal space is established for the dual coordinate ascent based stochastic algorithms. We also establish the accelerated linear complexity for some problems with nonsmooth loss, i.e., the least absolute deviation and SVM.

math.NA↗

arXiv2019

Hermitian matrices for clustering directed graphs: insights and applications

Graph clustering is a basic technique in machine learning, and has widespread applications in different domains. While spectral techniques have been successfully applied for clustering undirected graphs, the performance of spectral clustering algorithms for directed graphs (digraphs) is not in general satisfactory: these algorithms usually require symmetrising the matrix representing a digraph, and typical objective functions for undirected graph clustering do not capture cluster-structures in which the information given by the direction of the edges is crucial. To overcome these downsides, we propose a spectral clustering algorithm based on a complex-valued matrix representation of digraphs. We analyse its theoretical performance on a Stochastic Block Model for digraphs in which the cluster-structure is given not only by variations in edge densities, but also by the direction of the edges. The significance of our work is highlighted on a data set pertaining to internal migration in the United States: while previous spectral clustering algorithms for digraphs can only reveal that people are more likely to move between counties that are geographically close, our approach is able to cluster together counties with a similar socio-economical profile even when they are geographically distant, and illustrates how people tend to move from rural to more urbanised areas.

cs.LG↗

arXiv2019

Valence transition in topological Kondo insulator

We investigate the valence transition in three-dimensional topological Kondo insulator through slave-boson analysis of periodic Anderson model. By including the effect of intra-atomic Coulomb correlation $U_{fc}$ between conduction and local electrons, we find a first-order valence transition from Kondo region to mixed valence upon ascending of local level above a critical $U_{fc}$, and this valence transition usually occurs very close to or simultaneously with a topological transition. Near the parameter region of zero-temperature valence transition, rise of temperature can generate a thermal valence transition from mixed valence to Kondo region, accompanied by a first-order topological transition. Remarkably, above a critical $U_{fc}$ which is considerable smaller than that generating paramagnetic valence transition, the original continuous antiferromagnetic transition is shifted to first order one, at which a discontinuous valence shift takes place. Upon increased $U_{fc}$, the paramagnetic valence transition approaches then converges with the first-order antiferromagnetic transition, leaving an significant valence shift on the magnetic boundary. The continuous antiferromagnetic transition, first-order antiferromagnetic transition, paramagnetic valence transition and topological transitions are all summarized in a global phase diagram. Our proposed exotic transition processes can help to understand the thermal valence variation as well as the valence shift around the pressure-induced magnetic transition in topological Kondo insulator candidates and in other heavy-fermion systems.

cond-mat.str-el↗

arXiv2019

Electromechanical Brillouin scattering in integrated optomechanical waveguides

In the well-known stimulated Brillouin scattering (SBS) process, spontaneous acoustic phonons in materials are stimulated by laser light and scatter the latter into a Stokes sideband. SBS becomes more pronounced in optical fibers and has been harnessed to amplify optical signals and even achieve lasing. Exploitation of SBS has recently surged on integrated photonics platforms as simultaneous confinement of photons and phonons in waveguides leads to drastically enhanced interaction. Instead of being optically stimulated, coherent phonons can also be electromechanically excited with very high efficiency as has been exploited in radiofrequency acoustic filters. Here, we demonstrate electromechanically excited Brillouin scattering in integrated optomechanical waveguides made of piezoelectric material aluminum nitride (AlN). Acoustic phonons of 16 GHz in frequency are excited with nanofabricated electromechanical transducers to scatter counter-propagating photons in the waveguide into a single anti-Stokes sideband. We show that phase-matching conditions of Brillouin scattering can be tuned by varying both the optical wavelength and the acoustic frequency to realize tunable single-sideband modulation. Combining Brillouin scattering photonics with nanoelectromechanical systems, our approach provides an efficient interface between microwave and optical photons that will be important for microwave photonics and potentially quantum transduction.

physics.optics↗

arXiv2019

Provable Accelerated Gradient Method for Nonconvex Low Rank Optimization

Optimization over low rank matrices has broad applications in machine learning. For large scale problems, an attractive heuristic is to factorize the low rank matrix to a product of two much smaller matrices. In this paper, we study the nonconvex problem $\min_{U\in\mathcal{R}^{n\times r}} g(U)=f(UU^T)$ under the assumptions that $f(X)$ is restricted $μ$-strongly convex and $L$-smooth on the set $\{X:X\succeq 0,rank(X)\leq r\}$. We propose an accelerated gradient method with alternating constraint that operates directly on the $U$ factors and show that the method has local linear convergence rate with the optimal dependence on the condition number of $\sqrt{L/μ}$. Globally, our method converges to the critical point with zero gradient from any initializer. Our method also applies to the problem with the asymmetric factorization of $X=\widetilde U\widetilde V^T$ and the same convergence result can be obtained. Extensive experimental results verify the advantage of our method.

math.NA↗

arXiv2019

Interpreting the star formation - extinction relation with MaNGA

We investigate the resolved relation between local extinction and star formation surface density within nearby star-forming galaxies selected from the MaNGA survey. Balmer decrement measurements imply an extinction of the Hα line emission which scales approximately linearly with the logarithm of the star formation surface density: $ A_{Hα} = 0.46 \log(Σ_{SFR}) + 1.53$. Secondary dependencies are observed such that, at a given $Σ_{SFR}$, regions of lower metallicity and/or enhanced Hα equivalent width (EW) suffer less obscuration than regions of higher metallicity and/or lower Hα EW. Spaxels lying above the mean relation also tend to belong to galaxies that are more massive, larger and viewed under higher inclination than average. We present a simple model in which the observed trends can be accounted for by a metallicity-dependent scaling between $Σ_{SFR}$ and $Σ_{dust}$ via a super-linear Kennicutt-Schmidt relation ($n_{KS} \sim 1.47$) and a dust-to-gas ratio which scales linearly with metallicity (DGR($Z_{\odot}$) = 0.01). The relation between the resulting total dust column and observed effective extinction towards nebular regions requires a geometry for the relative distribution of Hα emitting regions and dust that deviates from a uniform foreground screen and also from an entirely homogeneous mixture of dust and emitting sources. The best-fit model features an Hα EW and galactocentric distance dependent fraction of the dust mass in a clumpy foreground screen in front of a homogeneous mixture.

astro-ph.GA↗

arXiv2018

Accelerated Alternating Direction Method of Multipliers: an Optimal $O(1/K)$ Nonergodic Analysis

The Alternating Direction Method of Multipliers (ADMM) is widely used for linearly constrained convex problems. It is proven to have an $o(1/\sqrt{K})$ nonergodic convergence rate and a faster $O(1/K)$ ergodic rate after ergodic averaging, which may destroy the sparsity and low-rankness in sparse and low-rank learning, where $K$ is the number of iterations. In this paper, we modify the accelerated ADMM proposed in [Y. Ouyang, Y. Chen, G. Lan, and E. Pasiliao, An Accelerated Linearized Alternating Direction Method of Multipliers, SIAM J. on Imaging Sciences, 2015, 1588-1623] and give an $O(1/K)$ nonergodic convergence rate analysis, which satisfies $|F(x^K)-F(x^*)|\leq O(1/K)$, $\|Ax^K-b\|\leq O(1/K)$ and $x^K$ has a more favorable sparseness and low-rankness than the ergodic result. As far as we know, this is the first $O(1/K)$ nonergodic convergent ADMM type method for general linearly constrained convex problems. Moreover, we show that the lower complexity bound of ADMM type methods for the separable linearly constrained nonsmooth convex problems is $O(1/K)$, which means that our method is optimal.

math.NA↗

arXiv2018

Hermitian Laplacians and a Cheeger inequality for the Max-2-Lin problem

We study spectral approaches for the MAX-2-LIN(k) problem, in which we are given a system of $m$ linear equations of the form $x_i - x_j \equiv c_{ij}\mod k$, and required to find an assignment to the $n$ variables $\{x_i\}$ that maximises the total number of satisfied equations. We consider Hermitian Laplacians related to this problem, and prove a Cheeger inequality that relates the smallest eigenvalue of a Hermitian Laplacian to the maximum number of satisfied equations of a MAX-2-LIN(k) instance $\mathcal{I}$. We develop an $\widetilde{O}(kn^2)$ time algorithm that, for any $(1-\varepsilon)$-satisfiable instance, produces an assignment satisfying a $\left(1 - O(k)\sqrt{\varepsilon}\right)$-fraction of equations. We also present a subquadratic-time algorithm that, when the graph associated with $\mathcal{I}$ is an expander, produces an assignment satisfying a $\left(1- O(k^2)\varepsilon \right)$-fraction of the equations. Our Cheeger inequality and first algorithm can be seen as generalisations of the Cheeger inequality and algorithm for MAX-CUT developed by Trevisan.

cs.DS↗

arXiv2018

Spectral Subspace Sparsification

We introduce a new approach to spectral sparsification that approximates the quadratic form of the pseudoinverse of a graph Laplacian restricted to a subspace. We show that sparsifiers with a near-linear number of edges in the dimension of the subspace exist. Our setting generalizes that of Schur complement sparsifiers. Our approach produces sparsifiers by sampling a uniformly random spanning tree of the input graph and using that tree to guide an edge elimination procedure that contracts, deletes, and reweights edges. In the context of Schur complement sparsifiers, our approach has two benefits over prior work. First, it produces a sparsifier in almost-linear time with no runtime dependence on the desired error. We directly exploit this to compute approximate effective resistances for a small set of vertex pairs in faster time than prior work (Durfee-Kyng-Peebles-Rao-Sachdeva '17). Secondly, it yields sparsifiers that are reweighted minors of the input graph. As a result, we give a near-optimal answer to a variant of the Steiner point removal problem. A key ingredient of our algorithm is a subroutine of independent interest: a near-linear time algorithm that, given a chosen set of vertices, builds a data structure from which we can query a multiplicative approximation to the decrease in the effective resistance between two vertices after identifying all vertices in the chosen set to a single vertex with inverse polynomial additional additive error in near-constant time.

cs.DS↗

arXiv2018

Optimization Algorithm Inspired Deep Neural Network Structure Design

Deep neural networks have been one of the dominant machine learning approaches in recent years. Several new network structures are proposed and have better performance than the traditional feedforward neural network structure. Representative ones include the skip connection structure in ResNet and the dense connection structure in DenseNet. However, it still lacks a unified guidance for the neural network structure design. In this paper, we propose the hypothesis that the neural network structure design can be inspired by optimization algorithms and a faster optimization algorithm may lead to a better neural network structure. Specifically, we prove that the propagation in the feedforward neural network with the same linear transformation in different layers is equivalent to minimizing some function using the gradient descent algorithm. Based on this observation, we replace the gradient descent algorithm with the heavy ball algorithm and Nesterov's accelerated gradient descent algorithm, which are faster and inspire us to design new and better network structures. ResNet and DenseNet can be considered as two special cases of our framework. Numerical experiments on CIFAR-10, CIFAR-100 and ImageNet verify the advantage of our optimization algorithm inspired structures over ResNet and DenseNet.

cs.CV↗

arXiv2018

Magnetic and topological transitions in three-dimensional topological Kondo insulator

By using an extended slave-boson method, we draw a global phase diagram summarizing both magnetic phases and paramagnetic (PM) topological insulating phases (TI$_s$) in three-dimensional topological Kondo insulator (TKI). By including electron hopping (EH) up to third neighbor, we identify four strong topological insulating (STI) phases and two weak topological insulating (WTI) phases, then the PM phase diagrams characterizing topological transitions between these TI$_s$ are depicted as functions of EH, $f$-electron energy level and hybridization constant. We also find an insulator-metal transition from a STI phase which has surface Fermi rings and spin textures in qualitative agreement to TKI candidate SmB$_6$. In weak hybridization regime, antiferromagnetic (AF) order naturally arises in the phase diagrams, and depending on how the magnetic boundary crosses the PM topological transition lines, AF phases are classified into AF topological insulator (AFTI) and non-topological AF insulator (nAFI), according to their $\mathcal{Z}_2$ indices. In two small regions of parameter space, two distinct topological transition processes between AF phases occur, leading to two types of AFTI, showing distinguishable surface dispersions around their Dirac points.

cond-mat.str-el↗

arXiv2018

$\mathcal{Z}_2$ classification for a novel antiferromagnetic topological insulating phase in three-dimensional topological Kondo insulator

Antiferromagnetic topological insulator (AFTI) is a topological matter that breaks time-reversal symmetry. Since its proposal, explorations of AFTI in strong-correlated systems are still lacking. In this paper, we show for the first time that a novel AFTI phase can be realized in three-dimensional topological Kondo insulator (TKI). In a wide parameter region, the ground states of TKI undergo a second-order transition to antiferromagnetic insulating phases which conserve a combined symmetry of time reversal and a lattice translation, allowing us to derive a $\mathcal{Z}_2$-classification formula for these states. By calculating the $\mathcal{Z}_2$ index, the antiferromagnetic insulating states are classified into (AFTI) or non-topological antiferromagnetic insulator (nAFI) in different parameter regions. On the antiferromagnetic surfaces in AFTI, we find topologically protected gapless Dirac cones inside the bulk gap, leading to metallic Fermi rings exhibiting helical spin texture with weak spin-momentum locking. Depending on model parameters, the magnetic transitions take place either between AFTI and strong topological insulator, or between nAFI and weak topological insulator. By varying some model parameters, we find a topological transition between AFTI and nAFI, driving by closing of bulk gap. Our work may account for the pressure-induced magnetism in TKI compound SmB$_6$, and helps to explore richer AFTI phases in heavy-fermion systems as well as in other strong-correlated systems.

cond-mat.str-el↗

arXiv2018

Maximizing the Number of Spanning Trees in a Connected Graph

We study the problem of maximizing the number of spanning trees in a connected graph by adding at most $k$ edges from a given candidate edge set. We give both algorithmic and hardness results for this problem: - We give a greedy algorithm that, using submodularity, obtains an approximation ratio of $(1 - 1/e - ε)$ in the exponent of the number of spanning trees for any $ε> 0$ in time $\tilde{O}(m ε^{-1} + (n + q) ε^{-3})$, where $m$ and $q$ is the number of edges in the original graph and the candidate edge set, respectively. Our running time is optimal with respect to the input size up to logarithmic factors, and substantially improves upon the $O(n^3)$ running time of the previous proposed greedy algorithm with approximation ratio $(1 - 1/e)$ in the exponent. Notably, the independence of our running time of $k$ is novel, comparing to conventional top-$k$ selections on graphs that usually run in $Ω(mk)$ time. A key ingredient of our greedy algorithm is a routine for maintaining effective resistances under edge additions in an online-offline hybrid setting. - We show the exponential inapproximability of this problem by proving that there exists a constant $c > 0$ such that it is NP-hard to approximate the optimum number of spanning trees in the exponent within $(1 - c)$. This inapproximability result follows from a reduction from the minimum path cover in undirected graphs, whose hardness again follows from the constant inapproximability of the Traveling Salesman Problem (TSP) with distances 1 and 2. Thus, the approximation ratio of our algorithm is also optimal up to a constant factor in the exponent. To our knowledge, this is the first hardness of approximation result for maximizing the number of spanning trees in a graph, or equivalently, by Kirchhoff's matrix-tree theorem, maximizing the determinant of an SDDM matrix.

cs.DS↗

arXiv2018

Optimized Projections for Compressed Sensing via Direct Mutual Coherence Minimization

Compressed Sensing (CS) is a novel technique for simultaneous signal sampling and compression based on the existence of a sparse representation of signal and a projected dictionary $PD$, where $P\in\mathbb{R}^{m\times d}$ is the projection matrix and $D\in\mathbb{R}^{d\times n}$ is the dictionary. To exactly recover the signal with a small number of measurements $m$, the projected dictionary $PD$ is expected to be of low mutual coherence. Several previous methods attempt to find the projection $P$ such that the mutual coherence of $PD$ can be as low as possible. However, they do not minimize the mutual coherence directly and thus their methods are far from optimal. Also the solvers they used lack of the convergence guarantee and thus there has no guarantee on the quality of their obtained solutions. This work aims to address these issues. We propose to find an optimal projection by minimizing the mutual coherence of $PD$ directly. This leads to a nonconvex nonsmooth minimization problem. We then approximate it by smoothing and solve it by alternate minimization. We further prove the convergence of our algorithm. To the best of our knowledge, this is the first work which directly minimizes the mutual coherence of the projected dictionary with a convergence guarantee. Numerical experiments demonstrate that the proposed method can recover sparse signals better than existing methods.

cs.IT↗

arXiv2018

Independence number and the number of maximum independent sets in pseudofractal scale-free web and Sierpiński gasket

As a fundamental subject of theoretical computer science, the maximum independent set (MIS) problem not only is of purely theoretical interest, but also has found wide applications in various fields. However, for a general graph determining the size of a MIS is NP-hard, and exact computation of the number of all MISs is even more difficult. It is thus of significant interest to seek special graphs for which the MIS problem can be exactly solved. In this paper, we address the MIS problem in the pseudofractal scale-free web and the Sierpiński gasket, which have the same number of vertices and edges. For both graphs, we determine exactly the independence number and the number of all possible MISs. The independence number of the pseudofractal scale-free web is as twice as the one of the Sierpiński gasket. Moreover, the pseudofractal scale-free web has a unique MIS, while the number of MISs in the Sierpiński gasket grows exponentially with the number of vertices.

cs.DS↗

arXiv2018

Current Flow Group Closeness Centrality for Complex Networks

Current flow closeness centrality (CFCC) has a better discriminating ability than the ordinary closeness centrality based on shortest paths. In this paper, we extend this notion to a group of vertices in a weighted graph, and then study the problem of finding a subset $S$ of $k$ vertices to maximize its CFCC $C(S)$, both theoretically and experimentally. We show that the problem is NP-hard, but propose two greedy algorithms for minimizing the reciprocal of $C(S)$ with provable guarantees using the monotoncity and supermodularity. The first is a deterministic algorithm with an approximation factor $(1-\frac{k}{k-1}\cdot\frac{1}{e})$ and cubic running time; while the second is a randomized algorithm with a $(1-\frac{k}{k-1}\cdot\frac{1}{e}-ε)$-approximation and nearly-linear running time for any $ε> 0$. Extensive experiments on model and real networks demonstrate that our algorithms are effective and efficient, with the second algorithm being scalable to massive networks with more than a million vertices.

cs.DS↗

arXiv2018

Kirchhoff Index As a Measure of Edge Centrality in Weighted Networks: Nearly Linear Time Algorithms

Most previous work of centralities focuses on metrics of vertex importance and methods for identifying powerful vertices, while related work for edges is much lesser, especially for weighted networks, due to the computational challenge. In this paper, we propose to use the well-known Kirchhoff index as the measure of edge centrality in weighted networks, called $θ$-Kirchhoff edge centrality. The Kirchhoff index of a network is defined as the sum of effective resistances over all vertex pairs. The centrality of an edge $e$ is reflected in the increase of Kirchhoff index of the network when the edge $e$ is partially deactivated, characterized by a parameter $θ$. We define two equivalent measures for $θ$-Kirchhoff edge centrality. Both are global metrics and have a better discriminating power than commonly used measures, based on local or partial structural information of networks, e.g. edge betweenness and spanning edge centrality. Despite the strong advantages of Kirchhoff index as a centrality measure and its wide applications, computing the exact value of Kirchhoff edge centrality for each edge in a graph is computationally demanding. To solve this problem, for each of the $θ$-Kirchhoff edge centrality metrics, we present an efficient algorithm to compute its $ε$-approximation for all the $m$ edges in nearly linear time in $m$. The proposed $θ$-Kirchhoff edge centrality is the first global metric of edge importance that can be provably approximated in nearly-linear time. Moreover, according to the $θ$-Kirchhoff edge centrality, we present a $θ$-Kirchhoff vertex centrality measure, as well as a fast algorithm that can compute $ε$-approximate Kirchhoff vertex centrality for all the $n$ vertices in nearly linear time in $m$.

cs.DS↗

arXiv2017

Consensus in Self-similar Hierarchical Graphs and Sierpiński Graphs: Convergence Speed, Delay Robustness, and Coherence

The hierarchical graphs and Sierpiński graphs are constructed iteratively, which have the same number of vertices and edges at any iteration, but exhibit quite different structural properties: the hierarchical graphs are non-fractal and small-world, while the Sierpiński graphs are fractal and "large-world". Both graphs have found broad applications. In this paper, we study consensus problems in hierarchical graphs and Sierpiński graphs, focusing on three important quantities of consensus problems, that is, convergence speed, delay robustness, and coherence for first-order (and second-order) dynamics, which are, respectively, determined by algebraic connectivity, maximum eigenvalue, and sum of reciprocal (and square of reciprocal) of each nonzero eigenvalue of Laplacian matrix. For both graphs, based on the explicit recursive relation of eigenvalues at two successive iterations, we evaluate the second smallest eigenvalue, as well as the largest eigenvalue, and obtain the closed-form solutions to the sum of reciprocals (and square of reciprocals) of all nonzero eigenvalues. We also compare our obtained results for consensus problems on both graphs and show that they differ in all quantities concerned, which is due to the marked difference of their topological structures.

eess.SY↗