## 美国休斯顿大学彭积明教授 5月22日上午做学术报告

Topic:The Direct Extension of ADMM for Multi-block Convex Minimization Problems is Not Necessarily Convergent

Abstract:

The issue of how to solve generic non-convex QP has been a long standing challenge in optimization. Existing global algorithms usually refer to branch-and-bound or successive relaxation approaches whose running time are typically exponential in term of the number of variables. Moreover, it has been proved that even finding a local optimal solution to LCQP is NP-hard.
In this talk, we introduce a new design paradigm for LCQPs with a few negative eigenvalues that are known to be NP-hard. We first introduce a new class of Lagrangian functions that satisfy the KKT conditions automatically. By using the new Lagrangian function, we present an alternative update scheme to improve the objective function. We then characterize the accumulation point of the sequence. By integrating the new algorithm and other simple optimization techniques such as convex relaxation, line search and partitioning, we present a global algorithm to find the global optimal solution to the underlying LCQP and estimate its complexity. Promising numerical experiments will be reported as well.

## 美国斯坦福大学叶荫宇教授 4月4日上午做学术报告

Topic: Optimization with Uncertain, Online and Massive Data

Abstract:

We present several analytic models and computational algorithms dealing with online/dynamic, structured and/or massively distributed data. Specifically, we discuss ：
• Distributionally Robust Optimization Models, where many problems can be efficiently solved when the associated uncertain data possess no priori distributions;
• Near-Optimal Online Linear Programming Algorithms, where the matrix data is revealed column by column along with the objective function and a decision has to be made as soon as a variable arrives;
•Sparse regression with Non-convex Regularization, where we give sparse and structure characterizations for every KKT stationary solution of the problem;
• Alternating Direction Method of Multipliers (ADMM) for large-scale data, where we give an example to show that the direct extension of ADMM for three-block convex minimization problems is not necessarily convergent, and propose simple and effective convergent variants.

## 热烈祝贺袁亚湘院士当选巴西科学院通讯院士

2014年12月3日, 巴西科学院公布2014年巴西科学院院士增选结果。我中心袁亚湘院士当选巴西科学院通讯院士，成为我国大陆数学领域的第一位巴西科学院通讯院士。

## 优化中心12月13日邀请报告

Topic:  优化算法开源库搭建

Abstract:

Topic: Warmstarting the Homogeneous and Self-Dual Interior Point Method for Linear and Conic Quadratic Problems

Abstract:

We present two strategies for warmstarting primal-dual interior point methods for the homogeneous self-dual model when applied to mixed linear and quadratic conic optimization problems. Common to both strategies is their use of only the final (optimal) iterate of the initial problem and their negligible computational cost. This is a major advantage when compared to previously suggested strategies that require a pool of iterates from the solution process of the initial problem. Consequently our strategies are better suited for users who use optimization algorithms as black-box routines which usually only output the final solution. Our two strategies differ in that one assumes knowledge only of the final primal solution while the other assumes the avail-ability of both primal and dual solutions.We analyze the strategies and deduce conditions under which they result in improved theoretical worst-case complexity. We present extensive computational results showing work reductions when warmstarting compared to coldstarting in the range 30%{75% depending on the problem class and magnitude of the problem perturbation. The computational experiments thus substantiate that the warmstarting strategies are useful in practice.
Joint work with Anders Skajaa and Erling Andersen.

Topic: 计及风电功率波动影响的风电场集群无功电压协调优化控制策略

Abstract:

## 2014年数学规划应用与软件研讨会学术报告

Topic: A class of polynomially solvable 0-1 programming problems and applications

Slides: Modeling01规划2014-Jinxing Xie

Topic: Linearlized Alternating Directions Methods of Multipliers in Sparse Optimization

Abstract:  The problem to find sparse solutions has obtained much attention and well studied widely in the fields of signal processing, compressive sensing, machine learning, statistical inference and so on. The problem mainly formulated as an unconstrained optimization problem which aims to minimize the sum of a smooth function and a non-smooth regularized term, the $\ell_1$-norm regularization, TV regularization, and matrix nuclear norm regularization, for instance. In this talk, we review some recent developed linearized alternating directions methods of multipliers to such problem, give their convergence results, and show their practical performance experimentally.

Slides: Prof. XiaoYunHai Repot_CAS

Topic: 基于矩阵稀疏表示的指纹图像压缩编码算法及软件实现

Topic: 高精度移动通信网络优化规划平台

Abstract:  随着蜂窝移动通信网络的规模不断扩大，无线网络的优化变得日益复杂和困难，仅仅依靠传统的工程优化和规划经验很难提升网络的性能。我们利用数学优化方法，结合电磁波理论，得到了一种高精度场强预测方法，可利用天线参数和地理信息等数据，计算出各天线在每个地理位置的场强值。在此基础上，对于给定的网络性能指标，我们建立了移动通信无线网络天线参数调整的最优化模型，并设计了相应的求解算法，使用该模型和算法可计算出全网或局部网络的天线参数整体优化调整方案。我们把网络性能指标的计算和展示、天线的自动调整和优化等功能编程实现，形成了一套完整的软件系统。该软件可根据使用者的不同需求，设计不同的操作流程以及交互界面，给网络优化工程师提供很好的技术支持。

Topic: 张量的低秩逼近

Abstract:  近年来涌现出许多高维图像数据，其本质上是高阶张量，如果采用传统图像处理方法将张量数据转换为非常长的向量来实现，这样测量就需要非常大样本矩阵，从而带来了巨大的计算和存储负担，同时，也会丧失张量所固有的几何、统计以及非线性度量性质。因此，将传统的理论方法拓展到高阶张量情形成为了亟待解决又非常具有挑战的一个崭新研究课题。本报告首先介绍张量计算，特别是张量低秩逼近计算的国内外研究现状，然后汇报我们的研究团队在张量低秩逼近方面的研究工作，主要包括低秩张量的完备化方法、对称复张量的最佳复秩1逼近的计算方法。

Slides: 张量的低秩逼近-Minru Bai

Topic: 混合整数规划算法与软件

Abstract:  许多实际问题可以归结为混合整数规划问题。比如，火车或飞机调度、工人排班、生产计划、发电机组组合、通讯中设备选址等。常见的混合整数规划问题有指派问题，0-1背包问题、设备覆盖问题、旅行商问题等。混合整数规划常用的算法包括: 预处理、分支定界、割平面、启发式算法等。此方面的软件也有很多，常用的商业软件有: Cplex、Gurobi、Xpress-MP等；常见的非商业软件有: Bonmin、Scip、Knitro等。虽然算法与软件都已经比较成熟，仍有很多问题值得思考。比如，如何针对特殊问题设计算法，如何提高算法的计算效率等。

Slides: 混合正数规划的算法与软件-Prof.Dai

Topic: 优化算法开源库搭建

Abstract:

## 2014年数学规划应用与软件研讨会

2014年数学规划应用与软件研讨会定于2014年11月14-16日在北京实创西山科技培训中心举办。会议由中国运筹学会数学规划分会和中科院数学与系统科学研究院优化与应用研究中心联合主办。会议将对数学规划第一届C++培训班进行经验总结，并对合格的学员颁发结业证书。此外，会议拟展开一系列讨论，主要议题是:“数学规划应用和软件开发”，包括优化算法包的开发，张量算法包的开发，混合整数规划算法包的开发等。会议将邀请袁亚湘院士莅临指导，邀请优化与应用领域的专家作报告，同时欢迎大家就目前各自感兴趣的问题进行研讨。

l  11月 14日(星期五)

1. 13:00-14:00在中科院数学与系统科学研究院南楼二楼报到；
2. 14:00: 数学规划应用与软件系列学术报告(详情另行通知)

l  11月15日(星期六) 上午8:00在中科院数学与系统科学院思源楼门口上车，一同前往北京实创西山科技培训中心。学术研讨会于下午2:00开始.学术议程如下(摘要见附件2).

1. 14:00-14:40  姜志鹏(中国科学院大学): 高精度移动通信网络优化规划平台
2. 14:40-15:20 白敏茹(湖南大学): 张量的低秩逼近
3. 15:20-16:00 戴彧虹(中科院数学与系统科学院): 混合整数规划算法与软件
4. 16:00-16:20  茶歇
5. 16:20-17:00 杨周旺(中国科学技术大学): 优化算法开源库搭建
6. 17:00-17:40  培训介绍，培训总结与颁发证书；讨论如何发展优化软件

l  11月16日(星期日)

2014年数学规划应用与软件研讨会报名表

## Congratulate YUAN Yaxiang won the TWAS Prizes for 2014 at the Academy’s 25th General Meeting in Muscat, Oman

TWAS Prizes 2014 are awarded in nine fields: Agricultural Sciences; Biology; Chemistry; Earth Sciences; Engineering Sciences; Mathematics; Medical Sciences; Physics; and Social Sciences (called the TWAS-Celso Furtado Prize). This year, there are 11 prize winners: two from Argentina; four from China; three from Taiwan, China; one from India and one from Malaysia. The prize winners include two women.

Each TWAS Prize carries a cash award of USD15,000. The winners will lecture about their research at TWAS’s 26th General Meeting in 2015, where they will also receive a plaque and the prize money.

Agricultural Sciences

ZHANG Fusuo of China, for establishing the principle and technology of integrated nutrient management in China that has helped enhance food production and nutrient use efficiency while reducing environmental damages nationwide.

Biology

Marcelo RUBINSTEIN of Argentina, for his significant contributions to the understanding of brain genes involved in appetitive behaviours, addiction and obesity using transgenic mouse technology.

Chemistry

XIE Yi of China, for her outstanding contribution in inorganic solid state chemistry at nanoscale, especially in inorganic functional solids with modulated electron and phonon structures.

Earth Sciences

CHUNG Sun-Lin of Taiwan, China, for his contribution to our understanding of continental tectonics and magma generation based on his studies from the roof of the world and neighbouring regions.

Engineering Sciences (shared)

• Viswanathan KUMARAN of India, for his path-breaking and all-encompassing work on the laminar-turbulent transition of flow through soft-walled tubes and channels, including theoretical prediction, experimental discovery and demonstration of practical applicability.
• LU Chih-Yuan of Taiwan, China, for his seminal scientific innovations and outstanding technical contributions in semiconductor device physics and semiconductor integrated circuits technology.

Mathematics

YUAN Yaxiang of China, for his contribution to numerical methods for nonlinear optimization, particularly to nonlinear conjugate gradient methods, trust region algorithms, quasi-Newton methods and subspace methods.

Medical Sciences (shared)

• Irene Oi Lin NG of China, for her contribution to our understanding of the molecular pathogenesis of liver cancer based on her novel findings in liver cancer stem cells, cell signaling pathways and molecular pathology.
• Tse Wen CHANG of Taiwan, China, for his invention and development of anti-IgE therapy (Xolair) used worldwide for treating severe persistent asthma untreatable with other medicine.

Physics

Daniel DE FLORIAN of Argentina, for his fundamental contribution to the understanding of Higgs Boson production at hadronic colliders.

TWAS-Celso Furtado Prize in Social Sciences

Rajah RASIAH of Malaysia, for his pioneering contribution to the concept and methodology of technological capabilities and establishing its relationship with economic development.

## Professor Jingyun Yuan from Federal University of Parana, Brazil will be invited to give a talk

Title:

Some Research Methodologies in Applied and Computational Mathematics

## 2014年“优化与应用”学术研讨会会议报告四

Topic:  Implementation of derivative-free optimization algorithms on land subsidence control problems

Abstract:  Due to surface water scarcity and contamination problems, groundwater has been exploited as an important water resource in many areas. However, the extensive pumping causes severe problems. In the Hang Jia Hu plain, land subsidence induced by groundwater extraction from a large number of pumping wells causes flooding, disrupted river navigation and waterlogging of soil, among other problems. In order to mitigate the hazards caused by land subsidence and also meet the domestic and industrial demands for water resources, groundwater exploitation management needs to reply on optimization techniques to determine the most effective strategies. However, this complex groundwater management problem is challenging, as its objective functions being nonsmooth, nonlinear, and having many local minima and lacking derivative forms. In our study, we apply the DYCORS algorithm, which is a derivative free optimization algorithm for solving expensive black-box objective functions. It is based on the radial basis function (RBF) surrogate model and combined with dynamic coordinate search so it is specifically suitable for high dimensional problems. With this optimization and simulation system, we will be able to provide efficient groundwater exploitation management plans under different practical requirements.

## 2014年“优化与应用”学术研讨会会议报告三

Topic:  一个新的非线性规划内点方法

Abstract: 内点方法是求解非线性规划的一类重要方法。传统的内点方法是通过求解松弛KKT条件的拟牛顿方程来产生迭代搜索方向，并通过选取适当的步长来保证迭代点总是内点。这样做有三个明显的缺陷：一是求解半定规划时需要使用对称化技术，导致算法十分复杂；二是为了保证内点可导致一个好的迭代步被截断；三是对于不满足MFCQ或LICQ的问题，Lagrange乘子估计会是无界的。我们提出一类新的内点方法，每次迭代求解一个带有两个参数的原始对偶方程，并能克服上述缺陷。我们将给出算法的全局和局部收敛性分析，并讨论一些相关的话题。一个数值例子表明对于不满足MFCQ或LICQ的问题我们能够获得较高精度的解。