arXiv subjects

Chao Xu

Publications and source records attributed to Chao Xu.

At least 379 records · Page 21Linked to original sources

arXiv2018

AutoEncoder Inspired Unsupervised Feature Selection

High-dimensional data in many areas such as computer vision and machine learning tasks brings in computational and analytical difficulty. Feature selection which selects a subset from observed features is a widely used approach for improving performance and effectiveness of machine learning models with high-dimensional data. In this paper, we propose a novel AutoEncoder Feature Selector (AEFS) for unsupervised feature selection which combines autoencoder regression and group lasso tasks. Compared to traditional feature selection methods, AEFS can select the most important features by excavating both linear and nonlinear information among features, which is more flexible than the conventional self-representation method for unsupervised feature selection with only linear assumptions. Experimental results on benchmark dataset show that the proposed method is superior to the state-of-the-art method.

cs.LG↗

arXiv2017

A Quantized Inter-level Character in Quantum Systems

For a quantum system subject to external parameters, the Berry phase is an intra-level property, which is gauge invariant module $2π$ for a closed loop in the parameter space and generally is non-quantized. In contrast, we define a inter-band character $Θ$ for a closed loop, which is gauge invariant and quantized as integer values. It is a quantum mechanical analogy of the Euler character based on the Gauss-Bonnet theorem for a manifold with a boundary. The role of the Gaussian curvature is mimicked by the difference between the Berry curvatures of the two levels, and the counterpart of the geodesic curvature is the quantum geometric potential which was proposed to improve the quantum adiabatic condition. This quantized inter-band character is also generalized to quantum degenerate systems.

cond-mat.quant-gas↗

arXiv2017

Towards Evolutional Compression

Compressing convolutional neural networks (CNNs) is essential for transferring the success of CNNs to a wide variety of applications to mobile devices. In contrast to directly recognizing subtle weights or filters as redundant in a given CNN, this paper presents an evolutionary method to automatically eliminate redundant convolution filters. We represent each compressed network as a binary individual of specific fitness. Then, the population is upgraded at each evolutionary iteration using genetic operations. As a result, an extremely compact CNN is generated using the fittest individual. In this approach, either large or small convolution filters can be redundant, and filters in the compressed network are more distinct. In addition, since the number of filters in each convolutional layer is reduced, the number of filter channels and the size of feature maps are also decreased, naturally improving both the compression and speed-up ratios. Experiments on benchmark deep CNN models suggest the superiority of the proposed algorithm over the state-of-the-art compression methods.

stat.ML↗

arXiv2017

Global and fixed-terminal cuts in digraphs

The computational complexity of multicut-like problems may vary significantly depending on whether the terminals are fixed or not. In this work we present a comprehensive study of this phenomenon in two types of cut problems in directed graphs: double cut and bicut. 1. The fixed-terminal edge-weighted double cut is known to be solvable efficiently. We show a tight approximability factor of $2$ for the fixed-terminal node-weighted double cut. We show that the global node-weighted double cut cannot be approximated to a factor smaller than $3/2$ under the Unique Games Conjecture (UGC). 2. The fixed-terminal edge-weighted bicut is known to have a tight approximability factor of $2$. We show that the global edge-weighted bicut is approximable to a factor strictly better than $2$, and that the global node-weighted bicut cannot be approximated to a factor smaller than $3/2$ under UGC. 3. In relation to these investigations, we also prove two results on undirected graphs which are of independent interest. First, we show NP-completeness and a tight inapproximability bound of $4/3$ for the node-weighted $3$-cut problem. Second, we show that for constant $k$, there exists an efficient algorithm to solve the minimum $\{s,t\}$-separating $k$-cut problem. Our techniques for the algorithms are combinatorial, based on LPs and based on enumeration of approximate min-cuts. Our hardness results are based on combinatorial reductions and integrality gap instances.

cs.DS↗

arXiv2017

Computing minimum cuts in hypergraphs

We study algorithmic and structural aspects of connectivity in hypergraphs. Given a hypergraph $H=(V,E)$ with $n = |V|$, $m = |E|$ and $p = \sum_{e \in E} |e|$ the best known algorithm to compute a global minimum cut in $H$ runs in time $O(np)$ for the uncapacitated case and in $O(np + n^2 \log n)$ time for the capacitated case. We show the following new results. 1. Given an uncapacitated hypergraph $H$ and an integer $k$ we describe an algorithm that runs in $O(p)$ time to find a subhypergraph $H'$ with sum of degrees $O(kn)$ that preserves all edge-connectivities up to $k$ (a $k$-sparsifier). This generalizes the corresponding result of Nagamochi and Ibaraki from graphs to hypergraphs. Using this sparsification we obtain an $O(p + λn^2)$ time algorithm for computing a global minimum cut of $H$ where $λ$ is the minimum cut value. 2. We generalize Matula's argument for graphs to hypergraphs and obtain a $(2+ε)$-approximation to the global minimum cut in a capacitated hypergraph in $O(\frac{1}ε (p \log n + n \log^2 n))$ time. 3. We show that a hypercactus representation of all the global minimum cuts of a capacitated hypergraph can be computed in $O(np + n^2 \log n)$ time and $O(p)$ space. We utilize vertex ordering based ideas to obtain our results. Unlike graphs we observe that there are several different orderings for hypergraphs which yield different insights.

cs.DS↗

arXiv2017

A Logic of Knowing Why

When we say "I know why he was late", we know not only the fact that he was late, but also an explanation of this fact. We propose a logical framework of "knowing why" inspired by the existing formal studies on why-questions, scientific explanation, and justification logic. We introduce the Ky_i operator into the language of epistemic logic to express "agent i knows why phi" and propose a Kripke-style semantics of such expressions in terms of knowing an explanation of phi. We obtain two sound and complete axiomatizations w.r.t. two different model classes depending on different assumptions about introspection.

cs.AI↗

arXiv2017

A note on approximate strengths of edges in a hypergraph

Let $H=(V,E)$ be an edge-weighted hypergraph of rank $r$. Kogan and Krauthgamer extended Benczúr and Karger's random sampling scheme for cut sparsification from graphs to hypergraphs. The sampling requires an algorithm for computing the approximate strengths of edges. In this note we extend the algorithm for graphs to hypergraphs and describe a near-linear time algorithm to compute approximate strengths of edges; we build on a sparsification result for hypergraphs from our recent work. Combined with prior results we obtain faster algorithms for finding $(1+ε)$-approximate mincuts when the rank of the hypergraph is small.

cs.DS↗

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

A Faster Pseudopolynomial Time Algorithm for Subset Sum

Given a multiset $S$ of $n$ positive integers and a target integer $t$, the subset sum problem is to decide if there is a subset of $S$ that sums up to $t$. We present a new divide-and-conquer algorithm that computes all the realizable subset sums up to an integer $u$ in $\widetilde{O}\!\left(\min\{\sqrt{n}u,u^{4/3},σ\}\right)$, where $σ$ is the sum of all elements in $S$ and $\widetilde{O}$ hides polylogarithmic factors. This result improves upon the standard dynamic programming algorithm that runs in $O(nu)$ time. To the best of our knowledge, the new algorithm is the fastest general algorithm for this problem. We also present a modified algorithm for cyclic groups, which computes all the realizable subset sums within the group in $\widetilde{O}\!\left(\min\{\sqrt{n}m,m^{5/4}\}\right)$ time, where $m$ is the order of the group.

cs.DS↗

arXiv2016

Dynamic Optimization of Trajectory for Ramp-up Current Profile in Tokamak Plasmas

In this paper, we consider an open-loop, finite-time, optimal control problem of attaining a specific desired current profile during the ramp-up phase by finding the best open-loop actuator input trajectories. Average density, total power, and plasma current are used as control actuators to manipulate the profile shape in tokamak plasmas. Based on the control parameterization method, we propose a numerical solution procedure directly to solve the original PDE-constrained optimization problem using gradient-based optimization techniques such as sequential quadratic programming (SQP). This paper is aimed at proposing an effective framework for the solution of PDE-constrained optimization problem in tokamak plasmas. A more user-friendly and efficient graphical user interface (GUI) is designed in MATLAB and the numerical simulation results are verified to demonstrate its applicability. In addition, the proposed framework of combining existing PDE and numerical optimization solvers to solve PDE-constrained optimization problem has the prospective to target challenge advanced control problems arising in more general chemical engineering processes.

physics.plasm-ph↗

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

Fundamental Limits of Exciton-Exciton Annihilation for Light Emission in Transition Metal Dichalcogenide Monolayers

We quantitatively illustrate the fundamental limit that exciton-exciton annihilation (EEA) may impose to the light emission of monolayer transition metal dichalcogenide (TMDC) materials. The EEA in TMDC monolayers shows dependence on the interaction with substrates as its rate increases from 0.1 cm2/s (0.05 cm2/s) to 0.3 cm2/s (0.1 cm2/s) with the substrates removed for WS2 (MoS2) monolayers. It turns to be the major pathway of exciton decay and dominates the luminescence efficiency when the exciton density is beyond 1010 cm-2 in suspended monolayers or 1011 cm-2 in supported monolayers. This sets an upper limit on the density of injected charges in light emission devices for the realization of optimal luminescence efficiency. The strong EEA rate also dictates the pumping threshold for population inversion in the monolayers to be 12-18 MW/cm2 (optically) or 2.5-4x105 A/cm2 (electrically).

physics.optics↗

arXiv2016

A Gradient-based Kernel Optimization Approach for Parabolic Distributed Parameter Control Systems

This paper proposes a new gradient-based optimization approach for designing optimal feedback kernels for parabolic distributed parameter systems with boundary control. Unlike traditional kernel optimization methods for parabolic systems, our new method does not require solving non-standard Riccati-type or Klein-Gorden-type partial differential equations (PDEs). Instead, the feedback kernel is parameterized as a second-order polynomial whose coefficients are decision variables to be tuned via gradient-based dynamic optimization, where the gradient of the system cost functional (which penalizes both kernel and output magnitude) is computed by solving a so-called costate PDE instandard form. Special constraints are imposed on the kernel coefficients to ensure that, under mild conditions, the optimized kernel yields closed-loop stability. Numerical simulations demonstrate the effectiveness of the proposed approach.

math.OC↗

arXiv2016

B-spline Shape from Motion & Shading: An Automatic Free-form Surface Modeling for Face Reconstruction

Recently, many methods have been proposed for face reconstruction from multiple images, most of which involve fundamental principles of Shape from Shading and Structure from motion. However, a majority of the methods just generate discrete surface model of face. In this paper, B-spline Shape from Motion and Shading (BsSfMS) is proposed to reconstruct continuous B-spline surface for multi-view face images, according to an assumption that shading and motion information in the images contain 1st- and 0th-order derivative of B-spline face respectively. Face surface is expressed as a B-spline surface that can be reconstructed by optimizing B-spline control points. Therefore, normals and 3D feature points computed from shading and motion of images respectively are used as the 1st- and 0th- order derivative information, to be jointly applied in optimizing the B-spline face. Additionally, an IMLS (iterative multi-least-square) algorithm is proposed to handle the difficult control point optimization. Furthermore, synthetic samples and LFW dataset are introduced and conducted to verify the proposed approach, and the experimental results demonstrate the effectiveness with different poses, illuminations, expressions etc., even with wild images.

cs.CV↗

arXiv2015

Measurement of accelerator neutron radiation field spectrum by Extended Range Neutron Multisphere Spectrometers and unfolding program

This paper described a measurement of accelerator neutron radiation field at a transport beam line of Beijing-TBF. The experiment place was be selected around a Faraday Cup with a graphite target impacted by electron beam at 2.5GeV. First of all, we simulated the neutron radiation experiment by FLUKA. Secondly, we chose six appropriate ERNMS according to their neutron fluence response function to measure the neutron count rate. Then the U_M_G package program was be utilized to unfolding experiment data. Finally, we drew a comparison between the unfolding with the simulation spectrum and made an analysis about the result.

physics.acc-ph↗

arXiv2015

Computational Optimal Control of the Saint-Venant PDE Model Using the Time-scaling Technique

This paper proposes a new time-scaling approach for computational optimal control of a distributed parameter system governed by the Saint-Venant PDEs. We propose the time-scaling approach, which can change a uniform time partition to a nonuniform one. We also derive the gradient formulas by using the variational method. Then the method of lines (MOL) is applied to compute the Saint-Venant PDEs after implementing the time-scaling transformation and the associate costate PDEs. Finally, we compare the optimization results using the proposed time-scaling approach with the one not using it. The simulation result demonstrates the effectiveness of the proposed time-scaling method.

eess.SY↗