大会特邀报告：

• 聂飞平（西北工业大学）
  
  西北工业大学教授、博士生导师，清华大学自动化系博士；曾在新加坡南洋理工大学从事研究工作，之后在美国德州大学阿灵顿分校先后担任研究助理教授，研究副教授，研究教授；2015年入选中组部青年千人计划。主要研究兴趣为模式识别与机器学习中的理论和方法设计，并将所设计的方法成功应用于图像分割与标注、多媒体信息理解与检索、生物信息学等多个领域的实际问题中。已在PAMI、IJCV、Bioinformatics、ICML、NIPS、SIGKDD等国际顶尖期刊和会议上发表学术论文200余篇，其中CCF A类论文130余篇。据Google Scholar统计，论文总引用为7000余次，H指数为47。常年应邀担任相关领域顶级期刊和会议的审稿专家或程序委员会委员，并同时应邀担任IEEE Transactions on Neural Networks and Learning Systems、Information Science等多个国际一流SCI期刊的编委。
  
  报告题目：基于结构化图学习的高效聚类方法
  摘要：数据聚类是机器学习和数据挖掘研究中的一个基本问题。在数十年的研究中，已经提出了很多聚类方法，而基于图的聚类方法是其中最有效的方法之一。传统的图聚类方法需要用户事先给定一个图，然后采用松弛技巧将问题转化为一个可解的问题。由于一般的图不具有结构性，因此得到的解是连续的，需要利用离散技术得到最终的聚类结果，从而使得聚类结果十分依赖于初始化。针对这些问题，我们提出了一种结构化图学习方法，通过学习一个具有结构的图，使得我们可以直接得到聚类结果，不再依赖于初始化。该新方法具有性能优越，稳定等优点，并且其中的结构化图学习思想可以应用在其他基于图的机器学习方法中，具有很大的应用价值和启发性。
  
  
• 顾险峰（State University of New York at Stony Brook, 美国 ）
  
  顾险峰博士于清华大学计算机系获得计算机科学与技术学士学位，哈佛大学计算机科学硕士和博士学位，师从国际著名微分几何大师丘成桐院士，现于纽约州立大学石溪分校计算机科学系和应用数学系终身教授，清华大学丘成桐数学科学中心客座教授，大连理工大学海天学者，首都师范大学数字几何和成像实验室主任等。顾博士曾于2005年获得美国国家自然科学基金CAREER奖，2006年获得中国国家自然科学基金海外杰出青年学者奖，2013年第六届世界华人数学家大会晨兴应用数学金奖等。顾险峰教授和丘成桐先生，及其合作者共同创立了一门新兴的跨领域学科：计算共形几何。这门学科结合了现代几何和计算机科学，广泛应用于计算机图形学，计算机视觉，可视化，几何建模，网络和医学图像等领域。顾博士在数学、工程和医学领域的国际顶级杂志和会议发表论文270多篇。顾博士担任国际期刊《Graphical Models》，《IEEE Transaction on Computer Graphics and Visualizatiaon》的编辑和《Geometry，Imaging and Computation》的主编。其主要著作有《计算共形几何》、《离散曲面的变分原理》和《Ricci Flow for Shape Analysis and Surface Registration》等。顾博士获得多项国际专利， 其中虚拟肠镜专利以百万美元专利费转让给西门子和GE公司。
  
  报告题目：Geometric Interpretation to Generative Learning Model

  摘要： Generative models play important roles in statistical learning, such as Generative Adversarial Network (GAN) models, Variational Auto Encoders (VAE), which learn statistical distributions and generate random variables. Although the learning models are effective, the fundamental principles remain obscure and mysterious.

  In this talk, we interpret the generative models using optimal mass transportation theoretic framework, explain the implicit algorithms using explicit mathematical theorems. We attempt to answer the following questions: why learning a statistical distribution is much easier than learning a map ? Whether a neural network really learns or just remembers ? Why sometimes the learning model is easily be fooled ? How to compare the learning effectiveness of an implicit network model with the optimal mass transportation model
  

   

• 孙晓明（中国科学院计算技术研究所）
  
  

  中科院计算所研究员。05年毕业于清华大学计算机系，获博士学位。曾任清华大学高等研究院助理研究员、副研究员。主要研究方向理论计算机科学，曾获得首批自然科学基金优秀青年基金资助，入选万人计划首批青年拔尖人才，还曾获中国密码学会优秀青年奖、密码创新奖二等奖。目前担任中国密码学会青年工作委员会委员，CCF理论专委会副主任, 学工委委员，《JCST》,《软件学报》,《计算机研究与发展》，《ACM China Magazine》等杂志编委。

  报告题目: 精确量子算法
  

  摘要： 量子计算是一种利用了量子力学特性进行计算的新型计算模型，已经在多个计算问题上展示出了超越经典计算机的计算能力，特别是Shor所提出的多项式时间大整数分解量子算法，目前已知的大整数分解问题的经典算法都需要指数时间。在这一报告中我们将简要回顾量子算法的发展，并汇报我们最近在搜索问题的精确量子算法设计和分析方面的一些工作进展。与Shor算法、Grover算法不同，精确量子算法能够在有限步内终止，并以概率1输出正确结果。

青年学者邀请报告：

• 陈绍示（中国科学院数学与系统科学研究院）
  题目：D-finite functions: rationality and singularity analysis
  摘要：D-finite functions are solutions of systems of linear partial differential equations with polynomial coefficients of special type. This class of functions has been systematically investigated by R. Stanley in his book Enumerative Combinatorics (Volume II) as basic generating functions in combinatorics. In this talk, we will present some results related to the loindexcal and global studies of D-finite functions. Singularity analysis explores the local behavior of analytic functions. In the local aspect, we will study the apparent singularities of D-finite functions and the corresponding desingularization algorithms. In the global aspect, we prove that a multivariate D-finite power series with coefficients from a finite set is rational. This generalizes a rationality theorem of van der Poorten and Shparlinski in 1996. As an application, we will show how this result can be used to study the nonnegative integer points on algebraic varieties. This talk is based on my recent joint works with Jason P. Bell, Manuel Kauers, Ziming Li, Michael Singer and Yi Zhang.
  
  冯如勇（中国科学院数学与系统科学研究院）
  题目：Direct and inverse problems in difference Galois theory
  摘要：Difference Galois theory is a generalization of the usual Galois theory to linear difference equations.
  Direct and inverse problems are two fundamental problems in this theory. Direct problem asks how to calculate the Galois group of a given linear difference equation, and inverse problem concerns which groups appear as Galois groups of linear difference equations. In this talk, we will give a brief introduction to difference Galois theory and survey recent work on direct and inverse problems.
  
  雷娜（大连理工大学）
  题目：Automatic hexahedral mesh generation based on surface foliations
  摘要：For the purpose of isogeometric analysis, one of the most common ways is to construct structured hexahedral meshes, which have regular tensor product structure, and fit them by volumetric T-Splines. This theoretic work proposes a novel surface quadrilateral meshing method, colorable quad-mesh, which leads to the structured hexahedral mesh of the enclosed volume for high genus surfaces.
  The work proposes the following algorithm: the user first specifies a height parameter for each loop from an admisible curve system, which is a set of disjoint, simple loops automaticly computed on a high genus surface; a unique Strebel differential is computed with the combinatorial type and the heights prescribed by the user; the Strebel differential assigns a flat metric on the surface and decomposes the surface into cylinders; a quad-mesh is generated by splitting each cylinder into two quadrilaterals, followed by subdivision; the surface cylindrical decomposition is extended inward to produce a solid cylindrical decomposition of the volume; the hexahedral meshing is generated for each volumetric cylinder and then glued together to form a globally consistent hex-mesh.
  
  李新（中国科学技术大学）
  题目：AS++ T-splines and Iso-geometric Analysis: a perfect couple
  摘要：This talk will review the background of T-spline based isogeometric analysis. As the main motivation of construction design-through-analysis technology, we identify a new class of T-splines which have tighter relation with design and meanwhile maintain all the basic and good mathematical properties as B-splines and AS T-splines. We refer to them as AS++ T-splines, which prefectly couple with design and analysis.
  
  申立勇（中国科学院大学）
  题目： 基于μ基的有理张量积曲面的隐式化
  摘要：μ基方法是曲线曲面隐式化的重要方法之一,但是基于μ基结式的曲面隐式化仍然不完善，例如μ基结式常常含有除隐式方程之外的多余因子。因此我们研究了使得结式无多余因子的μ基存在性，并定义此类μ基为强μ基。通过计算分析表明，大多数曲面并不含有强μ基。针对一般μ基，我们分析发现，μ基结式的多余因子只能来自与张量参数表示的部分基点或者参数无穷远点。进一步，我们鉴定出各种多余因子的类型并且给出了它们的计算方法，除去多余因子即可得到隐式方程。
  
  李冬梅（湖南科技大学）
  题目：Groebner bases and the equivalence of polynomial matrices
  摘要：We investigate the following conjecture and open problem:
  (1) A valuation ring R is 1-Groebner if and only if R is an archimedean ring;
  (2) Is a 1-Groebner valuation ring coherent?
  We firstly prove that a zero-dimensional valuation ring is 1-Groebner. Furthermore, we give a positive answer to (1) and a negative answer to (2).
  Finally, we present some new results on the equivalence of polynomial matrices.