Loading...

Table of Content

    15 March 2022, Volume 26 Issue 1
    Mini-batch stochastic block coordinate descent algorithm
    Jia HU, Tiande GUO, Congying HAN
    2022, 26(1):  1-22.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.001
    Asbtract ( 2431 )   HTML ( 408)   PDF (986KB) ( 420 )  
    Figures and Tables | References | Related Articles | Metrics

    We study the mini-batch stochastic block coordinate descent algorithm (mSBD) for a class of problems, i.e., structured stochastic optimization problems (where "structured" means that feasible region of the problem has a block structure and the nonsmooth regularized part of the objective function is separable across the variable blocks), widely used in machine learning. We give the basic mSBD and its variant according to solving non-composite and composite problems respectively. For the noncomposite problem, we analyze the convergence properties of the algorithm without the assumption of uniformly bounded gradient variance. For the composite problem, we obtain the convergence of the algorithm without the usual Lipschitz gradient continuity assumption. Finally, we verify the effectiveness of mSBD by numerical experiments.

    Four-party evolutionary game for e-commerce ecosystem
    Xinxin WANG, Yukun CHENG, Xiaoming TIAN, Zhiqi XU, Jinmian CHEN
    2022, 26(1):  23-42.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.002
    Asbtract ( 2541 )   HTML ( 29)   PDF (1491KB) ( 431 )  
    Figures and Tables | References | Related Articles | Metrics

    Inspired by the frequent exposure of consumer rights protection issues in recent years, this paper aims to explore the impact of consumer behaviors on the strategic choices of other groups in the e-commerce ecosystem. For this purpose, we construct a four-party evolutionary game model, involving the government, e-commerce platforms, businesses and consumers. The evolutionary stable strategies under different conditions of the four-party evolutionary game are explored and the effect from different factors on the decision-makings of participants for the e-commerce ecosystem are also analyzed. Numerical analysis results show that consumers' complaint behaviors are conducive to promoting strict government supervision, strict management of e-commerce platforms and honest operation of merchants. Our work enriches an understanding of the governance for the the e-commerce ecosystem, and provides theoretical support for the related participants to make proper decisions in the e-commerce ecosystem.

    An equivalence condition for sparse optimization problem with non-negative constraints
    Yaxing LYU, Meijia HAN, Zilin HUANG, Wenxing ZHU
    2022, 26(1):  43-59.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.003
    Asbtract ( 2214 )   HTML ( 20)   PDF (874KB) ( 228 )  
    References | Related Articles | Metrics

    Weighted l1 minimization is one of the mainstream methods for sparse optimization. Since the l0 minimization problem is NP-hard, this paper studies the relationship between solutions of the l0 minimization problem and the weighted l1 minimization problem with non-negative constraints, and gives the definition of "s-weighted goodness" for the constraint matrix and the coefficient of the objective function. Through this definition, a sufficient condition is provided for a solution of the weighted l1 minimization problem to be a solution of the l0 minimization problem with non-negative constraints. Some results and proofs are provided from the perspective of monotonicity and robust stability. Further, this paper gives a sufficient condition for "s-weighted goodness", and shows lower and upper bounds of the sufficient condition, which are easy to check. Moreover, this paper also conducts an error analysis of the solution of the weighted l1 minimization problem.

    The constrained multi-sources eccentricity augmentation problems
    Jianping LI, Lijian CAI, Junran LICHEN, Pengxiang PAN
    2022, 26(1):  60-68.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.004
    Asbtract ( 2195 )   HTML ( 14)   PDF (797KB) ( 179 )  
    References | Related Articles | Metrics

    Given a weighted graph $G=(V, E; w, c)$ and a spanning subgraph $G_{1}=(V, E_{1})$ of $G$, where a set $S=\{s_{1}, s_{2}, \cdots, s_{k}\}$ of $k$ sources in $V$, a weight function $w: E\rightarrow \mathbb{R}^{+}$, a cost function $c: E\setminus E_{1}\rightarrow \mathbb{Z}^{+}$, and a positive integer $B$, we consider two kinds of the constrained multi-sources eccentricity augmentation problems as follows. (1) The constrained multi-sources minimum eccentricity augmentation problem (the CMS-Min-EA problem, for short) is asked to find a subset $E_{2}\subseteq E\setminus E_{1}$ with a constraint $c(E_{2})\leq B$, the objective is to minimize the minimum of the eccentricities of vertices of $S$ in the graph $G_{1}\cup E_{2}$, and (2) The constrained multi-sources maximum eccentricity augmentation problem (the CMS-Max-EA problem, for short) is asked to find a subset $E_{2}\subseteq E\setminus E_{1}$ with a constraint $c(E_{2})\leq B$, the objective is to minimize the maximum of the eccentricities of vertices of $S$ in the graph $G_{1}\cup E_{2}$. For the two problems mentioned-above, we design two fixed parameter tractable (FPT) constant-approximation algorithms to solve them, respectively.

    A survey on approximation algorithms for one dimensional bin packing
    Hua CHEN, Guochuan ZHANG
    2022, 26(1):  69-84.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.005
    Asbtract ( 2637 )   HTML ( 50)   PDF (946KB) ( 462 )  
    References | Related Articles | Metrics

    With the emergence of computational complexity theory, the study of approximation algorithms has gradually become an important field in combinatorial optimization since the 1970s. As one of the classic combinatorial optimization problems, bin packing has attracted great attention. It can be widely found in various resource allocation problems with capacity constraints. Its more and more important applications including logistics loading and material cutting aside, any theoretical breakthrough of bin packing algorithms is related to the development of the whole combinatorial optimization. At present, the research on approximation algorithms of bin packing is still popular. This paper briefly introduces the process of some classic Fit algorithms, analyzes the main ideas of approximation schemes based on linear programming relaxation, reviews the state of the art, and provides some suggestions for future research.

    Streaming algorithms for the maximization of submodular+supermodular functions with a cardinality constraint
    Yuefang LIAN, Zhenning ZHANG, Zhongrui ZHAO, Dingzhu DU
    2022, 26(1):  85-98.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.006
    Asbtract ( 2558 )   HTML ( 20)   PDF (928KB) ( 282 )  
    Figures and Tables | References | Related Articles | Metrics

    In this paper, wepropose streaming algorithms for the maximization ofsubmodular+supermodular functions with cardinality constraint, whichhas wide applications in data processing, machine learning andartificial intelligence. By means of the diminishing return ratioof the objective function, we design a one-pass sieve-streamingalgorithm and get the approximate ratio $\min\left\{\frac{(1-\varepsilon)\gamma}{2^{\gamma}}, 1-\frac{\gamma}{2^{\gamma}(1-\kappa^{g})^{2}}\right\}$. Numerical experiments show that the sieve-streaming algorithm is effective forthe BP-maximization problem and can guarantee the same result asthe greedy algorithm with less time if submodular function andsupermodular are in the same order of magnitude.

    The seeding algorithm for $\mu$-similar Bregman divergences $k$-means problem with penalties
    Wenjie LIU, Dongmei ZHANG, Peng ZHANG, Juan ZOU
    2022, 26(1):  99-112.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.007
    Asbtract ( 2020 )   HTML ( 11)   PDF (887KB) ( 159 )  
    Figures and Tables | References | Related Articles | Metrics

    The $k$-means problem is a classic NP-hard problem in clustering analysis. If somedata points are allowed not to be clustered but to pay some penalty, the $k$-means problem with penalty is induced. In this paper, weextend the $k$-means problem with penalty from Euclidean distance tothe more general $\mu $-similar Bregman divergences, and study theinitialization algorithm for the $\mu $-similar Bregman divergences $k$-means problem with penalty. The seeding algorithm is given wherethe approximation ratio is only related to $\mu $ and the ratio ofthe maximum value to the minimum value of the data point penalty.

    A local search analysis for the uniform capacitated $k$-means problem with penalty
    Jiachen JU, Qian LIU, Zhao ZHANG, Yang ZHOU
    2022, 26(1):  113-124.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.008
    Asbtract ( 2128 )   HTML ( 223)   PDF (902KB) ( 174 )  
    Figures and Tables | References | Related Articles | Metrics

    The classical $k$-means problem has numerous application in the real world. Given a set of data points $D$ in ${\mathbb R}^d$ and an integer $k$, the aim is to select $k$ points in ${\mathbb R}^d$ as a central set $S$, such that the sum of the square of the distance between each point and its closest center in $S$ is minimized. It is a NP-hard problem and has many generalizations, one of which is the uniform capacitated $k$-means problem with penalty. Compared with the classical $k$-means problem, the penalty implies that each data point has a penalty cost, when the distance from some data points to their nearest center is greater than the penalty cost, they contribute the penalty cost. As for the uniform capacity, it suggests that each center is connected at most $U$ times. For this variant problem, we design a local search algorithm which selects at most $(3+\alpha)k$ centers and reaches $\beta$-approximate, where $\alpha>34$, $\beta>\frac{\alpha+34}{\alpha-34}$.

    Online single machine scheduling problem with transportation
    Yinling WANG, Xin HAN, Xinxin SHAO
    2022, 26(1):  125-133.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.009
    Asbtract ( 2224 )   HTML ( 12)   PDF (779KB) ( 221 )  
    References | Related Articles | Metrics

    This paper studies the single machine online scheduling problem with transporters. The problem assumes that the jobs arrive online over time, and there is only a single transporter in the system, which transports up to $k$ jobs each time. Each job needs to be processed on a single machine, and then transported to the destination by the transporter. The objective of the problem is to minimize the makespan, that is, the shortest time for all jobs to be processed and transported to the destination. For this problem, this paper studies the agreeable model, and proposes an online algorithm with the competitive ratio of $\frac{\sqrt{5}+1}{2}$ based on the greedy strategy, in the end, the algorithm is proved to be the optimal online algorithm.

    Online scheduling on single batch machine with variable lookahead interval
    Libo WANG, Wenhua LI, Dan YU
    2022, 26(1):  134-140.  doi:10.15960/j.cnki.issn.1007-6093.2022.01.010
    Asbtract ( 2166 )   HTML ( 11)   PDF (757KB) ( 198 )  
    References | Related Articles | Metrics

    This paper investigates the online scheduling on an unbounded parallel-batch machine with variable lookahead interval to minimize makespan. Online means that jobs arrive over time. At any time $t$, an online algorithm can foresee the information of the jobs arriving in $(t, t+\Delta(t)]$, where $\Delta(t)=\beta p_{\max}(t)$ is the length of the lookahead interval which is not invariable, $p_{\max}(t)$ is the maximum processing time of the jobs arriving at or before $t$ and $\beta\in(0, 1)$ is a constant. On the general case of processing time, we give an online algorithm which is the best possible for 0<β≤1/6. When the processing times of all jobs are restricted in an interval, we give the best possible online algorithm for 0<β<1.