Loading...

Table of Content

    15 September 2018, Volume 22 Issue 3
    Margin transfer-based multi-view support vector machine
    TANG Jingjing, TIAN Yingjie
    2018, 22(3):  1-14.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.001
    Asbtract ( 944 )   PDF (2660KB) ( 248 )  
    Related Articles | Metrics

    The data obtained from multiple sources  or different feature subsets are called multi-view data. Multi-view learning is a machine learning research field that models on the knowledge from multiple views. Many research works have verified that the utilization of multiple views can significantly improve the prediction effect of the model, so that a lot of models and algorithms are proposed. Existing multi-view learning models mainly follow the consensus principle and the complementarity principle. A typical SVM-based multi-view learning model, SVM-2K, extends support vector machine (SVM) for multi-view learning by using the distance minimization version of Kernel Canonical Correlation Analysis (KCCA). However, SVM-2K cannot fully unleash the power of the complementary information among different feature views. In this paper, we propose a new margin transfer-based multi-view support vector machine model, termed as M^2SVM. This brings a new model that incorporates both principles for multi-view learning. Furthermore, we theoretically analyze the performance of M^2SVM from the viewpoint of the consensus principle. Comparisons with SVM-2K reveal that M^2SVM is more flexible and favorable than SVM-2K. Experimental results on 50 binary data sets demonstrate the effectiveness of the proposed method.

    Spectral HS projection algorithm for solving nonlinear monotone equations
    CHEN Xiangping
    2018, 22(3):  15-27.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.002
    Asbtract ( 1057 )   PDF (1297KB) ( 107 )  
    Related Articles | Metrics

    In this paper, based on the structures of spectral gradient method and HS conjugate gradient method, we propose a spectral HS projection algorithm for solving nonlinear monotone equations. The proposed algorithm inherits some advantages of spectral gradient method and conjugate gradient method such as low memory cost and simple calculation. Moreover, it does not need any derivative information, since it is very suitable to solve non-smoothly nonlinear monotone equations. Under some appropriate conditions, we prove the convergence of the proposed method, and show the efficiency of the proposed method by some numerical experiments.

    A study of online berth and quay crane integrated allocation problem with lookahead ability
    LI Ying, QIAO Longliang, ZHENG Feifeng
    2018, 22(3):  28-36.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.003
    Asbtract ( 870 )   PDF (1103KB) ( 180 )  
    Related Articles | Metrics

    This paper studies an online over-list model of berth and quay crane integrated allocation problem, and it focuses on the case with lookahead ability such that on the release of any request, an online player can foresee the next k(k \geq 2) requests. The objective is to minimize the maximum completion time of a request, i.e., the makespan. It is assumed that there are two different sized requests, a hybrid berth consisting of three discrete berths and four quay cranes. We first prove several lower bounds of competitive ratio for k \geq 2, and then present an optimal 7/6-competitive algorithm for the case with k=2. Numerical results further demonstrate that the proposed algorithm behaves well in average performance.

    On-line strategies for multi-period newsvendor problem with price quantity discount
    ZHANG Yong, ZHONG Huifen, ZHANG Weiguo, et al.
    2018, 22(3):  37-48.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.004
    Asbtract ( 868 )   PDF (785KB) ( 116 )  
    Related Articles | Metrics

    The price quantity discount can increase the order quantity, which is an important factor of inventory decision making. Particularly, price discount only occurs when the order quantity reaches a fixed level. This paper uses the Weak Aggregating Algorithm (WAA) advanced in computer science to study the multi-period newsvendor problem with this kind price quantity discount. WAA is an on-line sequential decision-making algorithm; its main advantage is that it does not make statistical assumption on future inputs, which overcomes the difficulties of having to make probability hypothesis on demand in newsvendor problem research. Mainly, this paper applies WAA to experts whose strategies are fixed order quantities to present explicit online strategy for the multi-period newsvendor problem with price quantity discount. The theoretical guarantee for the proposed online strategy is obtained compared with the best expert strategy. The salvage and shortage cost are further introduced to obtain extended online strategies and their theoretical results. The numerical examples are finally used to show the good competitive performance of the proposed online strategies.

    An approximation algorithm for the squared metric dynamic facility location problem
    JIANG Yanjun, XU Dachuan, ZHANG Dongmei
    2018, 22(3):  49-58.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.005
    Asbtract ( 1026 )   PDF (1189KB) ( 112 )  
    Related Articles | Metrics

    We consider a generalization of the classic single-period metric facility location problem, which is  the squared metric dynamic facility location problem. We utilize the primal-dual scheme to design a 9-approximation algorithm. Then the approximation ratio is improved to 2.606 by the greedy improvement technique. We also give some discussion for several  variants of the squared metric dynamic facility location problem in  future work.

    A sufficient descent conjugate gradient method for nonlinear unconstrained optimization problems
    2018, 22(3):  59-68.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.006
    Asbtract ( 1116 )   PDF (3343KB) ( 188 )  
    Related Articles | Metrics

    One of the widely used methods for solving large scale unconstrained optimization problems is the conjugate gradient method. In this paper, we propose a new nonlinear conjugate gradient method (CG), which satisfies the sufficient descent condition independent of any line search. We further establish global convergence theorem of the new  CG method. Finally, a large amount of numerical experiments are carried  out  and reported.  It shows that the proposed method has  an efficient computational performance.

    Smoothing cautious DPRP conjugate gradient method for solving a kind of special nonsmooth equations with max-type function
    SHAO Shuting, DU Shouqiang
    2018, 22(3):  69-78.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.007
    Asbtract ( 984 )   PDF (988KB) ( 106 )  
    Related Articles | Metrics

    In this paper, we consider the method for solving a kind of special nonsmooth equations with max-type function and present a smoothing cautious DPRP conjugate gradient method based on the transformation by the smoothing function of the max-type function and the absolute value functions. The global  convergence of the smoothing cautious DPRP conjugate gradient method is given under the general assumptions. Finally, the effectiveness of the given method is shown by the related numerical experiments.

    Optimal conditions for the lower semicontionuity of efficient solution mapping to parametric generalized set-vector equilibrium problems
    MENG Xudong, WANG Sanhua, GONG Xunhua
    2018, 22(3):  79-88.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.008
    Asbtract ( 784 )   PDF (492KB) ( 90 )  
    Related Articles | Metrics

    The lower semicontionuity of weak efficient solution and efficient solution mappings to a class of parametric generalized set-vector equilibrium problems in real Hausdorff topological vector spaces are studied. Under the condition of nearly cone-subconvexlike, scalar characterization of weak efficient solution is given by using the scalar method. Under some suitable assumptions, the lower semicontionuity theorem of weak efficient solution and efficient solution mappings to the parametric generalized set-vector equilibrium problems are gained.

    Adaptive online portfolio strategies design and analysis in nonstationary market
    YANG Xingyu, HE Jin'an, LAI Mingcong
    2018, 22(3):  89-98.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.009
    Asbtract ( 901 )   PDF (3204KB) ( 91 )  
    Related Articles | Metrics

    Considering the behavior of the stock market is nonstationary and thus earlier observations are less relevant to the current investment decision-making, we design online portfolio strategies only based on recent stock price data. Firstly, we design an online portfolio strategy which is the weighted average of the previous portfolio and the best constant rebalanced portfolio corresponding to recent stock price data of fixed length. Secondly, we design an adaptive online portfolio strategy by choosing the weights via online learning. We present numerical analysis by using real stock data, and the results illustrate that our strategies perform well, compared with benchmark strategies and existing online portfolio strategies.

    Parallel machine scheduling with machine and human eligibility restrictions
    ZHAO Xiaocheng, LI Dagang
    2018, 22(3):  99-108.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.010
    Asbtract ( 952 )   PDF (785KB) ( 105 )  
    Related Articles | Metrics

    This paper considers the problem of parallel machine scheduling where both machine and human are essential resources with eligibility restrictions, the objective is to minimize the makespan. We focus on the case of unit-length jobs. Based on max-flow model and binary search algorithm, the problem can be solved in polynomial time with the bound of O(n^{3}logn). We further present an O(n^{2}) effective heuristic based on dual danymic flexibility selection(DDFS) strategy, which can achieve close or exact solution to optimality.  

    Two-agent scheduling on an unbounded parallel-batching machine to minimize maximum cost and makespan
    HE Cheng, HAN Xinxin
    2018, 22(3):  109-116.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.011
    Asbtract ( 935 )   PDF (535KB) ( 71 )  
    Related Articles | Metrics

    There are two agents A and B with each having their own job sets. The jobs of a common agent can be  processed in a common batch. Moreover, each agent has an objective function to be minimized. This paper studies the two-agent scheduling problem on an unbounded parallel-batching machine to minimize maximum cost of agent A and makespan of agent B simultaneously. We present a polynomial-time algorithm for finding all Pareto optimal points of the problem.

    An online algorithm for hierarchical scheduling on two identical machines with rejection
    MIN Xiao, ZHU Junlei, LIU Jing
    2018, 22(3):  117-124.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.012
    Asbtract ( 788 )   PDF (511KB) ( 150 )  
    Related Articles | Metrics

     Two identical machines M_{1}, M_{2} with same speed but different processing capabilities labeled by their hierarchies are provided. M_{1} has the hierarchy 1 and that of M_2 is 2. Jobs come one by one over list. Each job has three parameters: size t_{j}, hierarchy g_{j}=1,2, penalty p_{j}. When a job arrives, which can be rejected by paying its penalty or be accepted and scheduled on some machine whose hierarchy does not exceed the job's hierarchy. In fact, the job with hierarchy 1 can only be assigned to M_{1}, but the job with hierarchy 2 can be assigned to M_{1} or M_{2} . Preemption is not allowed. The objective is to minimize the sum of makespan yielded by all accepted jobs and the total penalties of all rejected jobs. For this problem, we present an online algorithm with a competitive ratio 11/6 and a relative lower bound 5/3.

    A penalized FB function for symmetric cone complementarity problems
    GAO Leifu, ZHANG Yahong
    2018, 22(3):  125-131.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.013
    Asbtract ( 827 )   PDF (472KB) ( 76 )  
    Related Articles | Metrics

    With Euclidean Jordan algebras, we proved the level-boundedness of the merit function related to a penalized Fischer-Burmeister function for symmetric cone complementarity problems with monotonicity in a method of inner product.The method has more universality and promotion value both on theories and applications compared with previous trace inequality method to prove level-boundedness of the merit function. Level-boundedness plays an important part on a guarantee of decline algorithm convergence when we design algorithm to solve unconstrained minimization problem. Therefore, it has theoretical significance on the design of algorithm.

    Researching of the scattered layout problem in circular zone
    YU Shanen, XU Wenyang, LIU Guangyu
    2018, 22(3):  132-138.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.014
    Asbtract ( 846 )   PDF (2154KB) ( 111 )  
    Related Articles | Metrics

    In this paper, we studied the scattered layout problem in a circular zone by establishing a non-linear mathematical programming model, which can be solved by gradient method after transformed into the maximum value search for a no-constrained optimization problem when the number of layout points is less. However, the gradient method will become extremely inefficient or even can not work if the layout points are many. In order to solve this problem we proposed an approximation algorithm with the bound of 1/2 for fast solution, And several cases are given to verify the rationality and performance of the approximate algorithm at last. In a sense, the conclusions and approximation algorithm of this article enrich and improve the theories of the scattered layout problem.

    Equivalent Lipschitz optimization model for the group zero-norm regularized problem
    CHEN Xingwen, PAN Shaohua
    2018, 22(3):  139-144.  doi:10.15960/j.cnki.issn.1007-6093.2018.03.015
    Asbtract ( 865 )   PDF (505KB) ( 128 )  
    Related Articles | Metrics

    With the help of the variational characterization of the zero-norm function, we reformulate the group zero-norm regularized problem as a MPCC (mathematical program with a complementarity constraint) and show that the penalty problem, yielded by moving the complementarity constraint into the objective, is a global exact penalty of the MPCC problem itself. The objective function of the exact penalty problem is not only global Lipschitz continuous in the feasible set but also has the desired bilinear structure, thereby providing a favorable equivalent Lipschitz optimization model for designing sequential convex relaxation algorithms of the group zero-norm regularized problem.