具有线性化技术的三块非凸不可分优化问题BregmanADMM收敛性分析

交替方向乘子法是求解两块可分离凸优化问题的有效方法,但是对于三块不可分的非凸优化问题的交替方向乘子法的收敛性可能无法保证.该文主要研究的是用线性化广义Bregman交替方向乘子法(L-G-BADMM)求解目标函数是三块不可分的非凸极小化问题的收敛性分析.在适当假设条件下,对算法中子问题进行求解并构建满足Kurdyka-Lojasiewicz性质的效益函数,经过理论证明可以得到该算法的收敛性.     

一类自适应广义交替方向乘子法

广义交替方向乘子法是求解凸优化问题的有效算法.当实际问题中子问题难以求解时,可以采用在子问题中添加邻近项的方法处理,邻近矩阵正定时,算法收敛,然而这也会使迭代步长较小.最新研究表明,邻近矩阵可以有一定的不正定性.本文在基于不定邻近项的广义交替方向乘子法框架下,提出一种自适应的广义交替方向乘子法,动态地选择邻近矩阵,增大迭代步长.在一些较弱的假设下,证明了算法的全局收敛性.我们进行一些初等数值实验,验证了算法的有效性.     

修正乘子交替方向法求解三个可分离算子的凸优化

指出直接推广的经典乘子交替方向法对三个算子的问题不能保证收敛的原因, 并且给出将其改造成收敛算法的相应策略. 同时, 在一个统一框架下, 证明了修正的乘子交替方向法的收敛性和遍历意义下具有O(1/t)~收敛速率.     

基于交替方向乘子法的大规模线性多商品流问题求解算法

企业的商品流通配送问题是典型的线性多商品流问题.由于经营规模的扩大和全球化运营模式的推行,企业所面临的问题规模正变得空前巨大,数据存储也越来越分散,传统方法已无法适应求解需求.本文基于交替方向乘子法(ADMM)的可分解性,提出一类随机ADMM算法,将大规模的问题分解成多个、规模比较小的问题,并采取随机顺序去求解这些小问题以及对偶问题,最终得到原问题的最优解.算法克服了ADMM的直接拓展求解多块问题时可能发散的缺点,并采用MnetGen生成器随机生成的多个规模不同的线性多商品流问题对算法进行了测试,验证了算法的有效性和高效的求解效率.     

求解一类结构型变分不等式的加速随机方法

基于Glowinski的交替方向法和何炳生教授的改善步长的收缩算法,提出一个求解结构型变分不等式的加速随机方法.新方法的优势在于利用独立同分布的随机数来扩张步长,克服了传统的交替方向法中固定步长因子的缺点,证明了新方法的下降方向是可行的.在适当的假设条件下,给出新方法的性质,并证明新方法依概率收敛.通过对来自于金融和统计中问题的一系列数值试验,验证新方法的可行性和有效性.     

一种求解单调变分不等式的部分并行分裂LQP交替方向法

LQP交替方向法是求解可分离结构型单调变分不等式问题的一种非常有效的方法.它不仅可以充分地利用目标函数的可分结构,将原问题分解为多个更易求解的子问题,还更适合求解大规模问题.对于带有三个可分离算子的单调变分不等式问题,结合增广拉格朗日算法和LQP交替方向法提出了一种部分并行分裂LQP交替方向法,构造了新算法的两个下降方向,结合这两个下降方向得到了一个新的下降方向,沿着这个新的下降方向给出了最优步长.并在较弱的假设条件下,证明了新算法的全局收敛性.     

一种求解单调变分不等式的部分并行分裂LQP交替方向法

LQP交替方向法是求解可分离结构型单调变分不等式问题的一种非常有效的方法.它不仅可以充分地利用目标函数的可分结构,将原问题分解为多个更易求解的子问题,还更适合求解大规模问题.对于带有三个可分离算子的单调变分不等式问题,结合增广拉格朗日算法和LQP交替方向法提出了一种部分并行分裂LQP交替方向法,构造了新算法的两个下降方向,结合这两个下降方向得到了一个新的下降方向,沿着这个新的下降方向给出了最优步长.并在较弱的假设条件下,证明了新算法的全局收敛性.     

随机Bregman ADMM及其在训练具有离散结构的支持向量机中的应用

针对具有多块可分结构的非凸优化问题提出了一类新的随机Bregman交替方向乘子法,在周期更新规则下, 证明了该算法的渐进收敛性; 在随机更新的规则下, 几乎确定的渐进收敛性得以证明。数值实验结果表明, 该算法可有效训练具有离散结构的支持向量机。     

非空集类有生成半域的条件

探讨一个非空集类是否一定有生成半域的问题．通过类比的方法提出“非空集类的生成半域”的概念。反例说明并非任意的非空集类都有生成半域．给出非空集类有生成半域的一个充分条件,以及在该条件下非空集类的生成半域的构造方法．最后给出由任意的非空集类构造其生成域的办法．     

非凸优化带松弛步的对称ADMM局部线性收敛率分析

基于对称交替方向乘子法(ADMM),结合松弛步技巧,该文提出一种带松弛步的对称ADMM用于求解两分块线性约束非凸优化问题.同时,新算法乘子更新步采用不同的松弛因子.常规假设下,给出新算法子序列的收敛性证明.误差界条件下,分析并获得由新算法产生的迭代点列以线性收敛的速率局部趋于问题稳定点,相应增广拉格朗日函数序列亦线性收敛.最后,初步试验结果表明新算法是有效的.     

Beurling Type Theorem on the Hilbert Space Generated by a Positive Sequence

Let H²(γ) be the Hilbert space over the bidisk D² generated by a positive sequence γ={γ_nm}_n,m ≥ 0. In this paper, we prove that the Beurling type theorem holds for the shift operator on H²(γ) with γ={γ_nm}_n,m ≥ 0 satisfying certain series of inequalities. As a corollary, we give several applications to a class of classical analytic reproducing kernel Hilbert spaces over the bidisk D².     

Weak (1,1) Boundedness of Oscillatory Singular Integral with Variable Phase

In this paper, the weak (1, 1) boundedness of oscillatory singular integral with variable phase P (x)γ(y) for any x, y ∈ R,
Tf(x):=p. v.∫_-∞^∞ e^iP(x)γ(y)f(x-y) dy/y
is studied, where P is a real monic polynomial on R.     

Extensions of n-Hom Lie algebras

n-Hom Lie algebras are twisted by n-Lie algebras by means of twisting maps. n-Hom Lie algebras have close relationships with statistical mechanics and mathematical physics. The paper main concerns structures and representations of n-Hom Lie algebras. The concept of n_ρ-cocycle for an n-Hom Lie algebra (G, [,… , ], α) related to a G-module (V, ρ, β) is proposed, and a sufficient condition for the existence of the dual representation of an n-Hom Lie algebra is provided. From a G-module (V, ρ, β) and an n_ρ-cocycle θ, an n-Hom Lie algebra (T_θ(V ), [, … , ]θ, γ) is constructed on the vector space T_θ(V ) = G⊕V, which is called the T_θ-extension of an n-Hom Lie algebra (G, [, … , ], α) by the G-module (V, ρ, β).     

Global well-posedness of the 3D generalized rotating magnetohydrodynamics equations

In this paper,we establish the global well-posedness of the generalized rotating magnetohydrodynamics equations if the initial data are in X~(1-2α) defined by X~(1-2α)={u∈D'(R~3):∫_(R~3)|ξ|~(1-2α)|(ξ)|dξ+∞}.In addition,we also give Gevrey class regularity of the solution.     

Inequalities for the Extended Best Polynomial Approximation Operator in Orlicz Spaces

In this paper we pursue the study of the best approximation operator extended from L~Φ to L~φ, where φ denotes the derivative of the function Φ. We get pointwise convergence for the coefficients of the extended best approximation polynomials for a wide class of function f, closely related to the Calder′on–Zygmund class t_m~p(x) which had been introduced in 1961. We also obtain weak and strong type inequalities for a maximal operator related to the extended best polynomial approximation and a norm convergence result for the coefficients is derived. In most of these results, we have to consider Matuszewska–Orlicz indices for the function φ.     

<Emphasis Type="Italic">n</Emphasis>-transitivity of bisection groups of a Lie groupoid

The notion of n-transitivity can be carried over from groups of diffeomorphisms on a manifold M to groups of bisections of a Lie groupoid over M. The main theorem states that the n-transitivity is fulfilled for all n ∈ N by an arbitrary group of C^r-bisections of a Lie groupoid Γ of class C^r, where 1 ≤ r ≤ ω, under mild conditions. For instance, the group of all bisections of any Lie groupoid and the group of all Lagrangian bisections of any symplectic groupoid are n-transitive in the sense of this theorem. In particular, if Γ is source connected for any arrow γ ∈ Γ, there is a bisection passing through γ.     

Homogenization of Elliptic Problems with Neumann Boundary Conditions in Non-smooth Domains

We consider a family of second-order elliptic operators {L_ε} in divergence form with rapidly oscillating and periodic coefficients in Lipschitz and convex domains in R~n. We are able to show that the uniform W~(1,p) estimate of second order elliptic systems holds for 2n/(n+1)-δ p 2n/(n-1)+ δ where δ 0 is independent of ε and the ranges are sharp for n = 2, 3. And for elliptic equations in Lipschitz domains, the W~(1,p) estimate is true for 3/2-δ p 3 + δ if n ≥ 4, similar estimate was extended to convex domains for 1 p ∞.     

On the Characterization of Maximal Planar Graphs with a Given Signed Cycle Domination Number

Let G =(V, E) be a simple graph. A function f : E → {+1,-1} is called a signed cycle domination function(SCDF) of G if ∑_(e∈E(C))f(e) ≥ 1 for every induced cycle C of G. The signed cycle domination number of G is defined as γ'_(sc)(G) = min{∑_(e∈E)f(e)| f is an SCDF of G}. This paper will characterize all maximal planar graphs G with order n ≥ 6 and γ'_(sc)(G) = n.     

Irreducible Wakimoto-like Modules for the Lie Superalgebra <Emphasis Type="Italic">D</Emphasis>(2, 1;α)

By using the idea of Wakimoto's free field, we construct a class of representations for the Lie superalgebra D(2, 1; α) on the tensor product of a polynomial algebra and an exterior algebra involving one parameter λ. Then we obtain the necessary and sufficient condition for the representations to be irreducible. In fact, the representation is irreducible if and only if the parameter λ satisfies (λ + m)(λ-(1+α/α)≠m)=0 for any m ∈ Z₊.     

An efficient inexact symmetric Gauss–Seidel based majorized ADMM for high-dimensional convex composite conic programming

In this paper, we propose an inexact multi-block ADMM-type first-order method for solving a class of high-dimensional convex composite conic optimization problems to moderate accuracy. The design of this method combines an inexact 2-block majorized semi-proximal ADMM and the recent advances in the inexact symmetric Gauss–Seidel (sGS) technique for solving a multi-block convex composite quadratic programming whose objective contains a nonsmooth term involving only the first block-variable. One distinctive feature of our proposed method (the sGS-imsPADMM) is that it only needs one cycle of an inexact sGS method, instead of an unknown number of cycles, to solve each of the subproblems involved. With some simple and implementable error tolerance criteria, the cost for solving the subproblems can be greatly reduced, and many steps in the forward sweep of each sGS cycle can often be skipped, which further contributes to the efficiency of the proposed method. Global convergence as well as the iteration complexity in the non-ergodic sense is established. Preliminary numerical experiments on some high-dimensional linear and convex quadratic SDP problems with a large number of linear equality and inequality constraints are also provided. The results show that for the vast majority of the tested problems, the sGS-imsPADMM is 2–3 times faster than the directly extended multi-block ADMM with the aggressive step-length of 1.618, which is currently the benchmark among first-order methods for solving multi-block linear and quadratic SDP problems though its convergence is not guaranteed.