We survey some recent graph algorithms that are based on picking a vertex at random and declaring it to be a part of the solution. This simple idea has been deployed to obtain state-of-the-art parameterized, exact exponential time, and approximation algorithms for a number of problems, such as Feedback Vertex Set and 3-Hitting Set. We will also discuss a recent 2-approximation algorithm for Feedback Vertex Set in Tournaments that
is based on picking a vertex at random and declaring it to not be part of the solution.

Speaker Bio

Saket Saurabh received his PhD in Computer Science (2008), from The Institute of Mathematical Sciences (IMSc), Chennai. Saurabh spent two years (2007-2009) as a Postdoctoral Fellow at University of Bergen, Norway, and is now a professor at IMSc and at Department of Informatics at the University of Bergen. His main research interests are in graph algorithms, parameterized algorithms and complexity. He has written more than 300 articles, graduated more than 15 PhD students, and mentored around 10 post doctoral fellows. He is a co-author of two books: Parameterized Algorithms and Kernelization theory of parameterized preprocessing.

He is a SwarnaJayanti Fellow in Mathematical Sciences (2018), Fellow of Indian Academy of Sciences (2020), Academia Europaea (2020), and European Association for Theoretical Computer Science (EATCS, 2021). He received the inaugural ACM India Early Career Researcher Award in 2020, and Shanti Swarup Bhatnagar Prize (SSB) for Science and Technology 2021 (Mathematical Science). He is also the recipient of an ERC starting grant and an ERC Consolidator Grants in parameterized algorithms. He was also named one of the ACM Distinguished Members in 2022.

Poster

Courcelle’s theorem is one of the most celebrated algorithmic meta-theorems and has brought a profound impact on the theory of parameterized complexity. The theorem states that every graph property expressible by a monadic second-order (MSO) formula $\phi$ can be checked in $f(\phi,t)\cdot n$ time where $n$ and $t$ denote the number of vertices and the treewidth of the input graph, respectively. While the time complexity of Courcelle’s theorem is linear in $n$, the understanding of the function $f(\phi,t)$ in the bound remained poor and coarse. In this talk, we discuss a fine-grained version of Courcelle’s theorem with explicit dependence on the quantifier structure of $\phi$ and the treewidth parameter $t$, which is almost optimal assuming the ETH.

Speaker Bio

Jie Xue is an Assistant Professor of Computer Science at NYU Shanghai. His research area is Theoretical Computer Science, and more specifically, Computational Geometry, Algorithms & Data Structures, Graph Theory, and Parameterized Complexity. His research mainly focuses on designing efficient algorithms and data structures for fundamental problems regarding geometric objects and graphs. His work has been published regularly on top venues of TCS, such as STOC, FOCS, SODA, SoCG, etc. and has received the Best Paper Award at SoCG’24 and a Distinguished Paper Award at AAAI’23.

Large Language Models (LLMs) are transforming many applications, yet their use in industrial and high-stakes settings remains limited. LLMs often suffer from hallucinations, limited interpretability, weak explicit reasoning, and heavy data dependence, making it difficult to ensure reliable, fair, and robust decision-making in real time.

This talk advocates a multi-paradigm AI approach that integrates machine learning with logical reasoning. By embedding data-driven models within logical rules as predicates, this approach improves reasoning consistency, interpretability, and controllability while reducing reliance on large training data. We present practical case studies spanning industrial manufacturing, cyber security, early-stage drug discovery, intelligent recommendation, and banking risk control, demonstrating how multi-paradigm AI supports low-cost, high-accuracy, and interpretable decision-making. The talk aims to incite interest in this emerging direction.

Speaker Bio

Professor Wenfei Fan is the Chair of Web Data Management at the University of Edinburgh, UK, and the Chief Scientist of Shenzhen Institute of Computing Science, China. He is a Foreign Member of Chinese Academy of Sciences, a Fellow of the Royal Society (FRS), a Fellow of the Royal Academy of Engineering (FREng), a Fellow of the Royal Society of Edinburgh (FRSE), a Member of the Academy of Europe (MAE), and an ACM Fellow (FACM). He is a visiting chair professor at Peking University, and a distinguished visiting professor at Tsinghua University.
He received his PhD from the University of Pennsylvania (USA), and his MSc and BSc from Peking University (China). He is a recipient of Royal Society Wolfson Research Merit Award in 2018, ERC Advanced Grant in 2015, the Roger Needham Award in 2008 (UK), Yangtze River Scholar in 2007 (China), the Outstanding Overseas Young Scholar Award in 2003 (China), the Career Award in 2001 (USA), and several Test-of-Time and Best Paper Awards (Alberto O. Mendelzon Test-of-Time Award of ACM PODS 2015 and 2010, Best Paper Awards for SIGMOD 2017, VLDB 2010, ICDE 2007 and Computer Networks 2002). His current research interests include database theory and systems, in particular big data, DB4AI, AI4DB, data quality, parallel models, and unification of logical reasoning and machine learning.

We study the Pigeonhole Equal Subset Sum problem, which is a total-search variant of the Subset Sum problem introduced by Papadimitriou (1994): we are given a set of $n$ positive integers $\{w_1,\ldots,w_n\}$ with the additional restriction that $\sum_{i=1}^n w_i \lt 2^n-1$, and want to find two different subsets $A,B \subseteq [n]$ such that $\sum_{i\in A} w_i = \sum_{i\in B} w_i$.

Previously, the best known runtime for Pigeonhole Equal Subset Sum was $O^*(2^{n/2})$, via either meet-in-middle or dynamic programming (Allcock, Hamoudi, Joux, Klingelhöfer, and Santha, ESA 2022). Our main result is a faster randomized algorithm in $O^*(2^{n/3})$ time. Unlike many previous works in this area, our approach does not use the representation method, but rather exploits a simple structural characterization of input instances with few solutions.

Based on two papers which appeared in ICALP 2024 and ESA 2025, jointly with Hongxun Wu, Ryan Williams, and Stan Zhang.

Speaker Bio

Ce Jin is a Miller Postdoctoral Fellow at UC Berkeley. He completed his PhD at MIT in 2025. Before that, he was an undergraduate student in Yao Class, Tsinghua University. He has a broad interest in theoretical computer science.

布尔可满足性问题（SAT）是理论计算机的经典问题，同时有广泛的应用。SAT求解器是电子设计自动化（EDA）的重要基础引擎，尤其是现代EDA形式化验证技术的基础。本报告主要介绍我们在SAT求解器以及EDA形式化验证技术的研究进展。

Speaker Bio

蔡少伟，中国科学院软件研究所研究员，中国科学院优秀导师，CCF杰出会员/学术工委执行委员，研究约束求解和形式化验证，主持自然基金委青年B项目和重点项目，获得领域顶级会议CAV、CP、SAT等会议的最佳/杰出论文奖，多次获得SAT比赛、SMT比赛和MaxSAT比赛冠军，以及鸿蒙创新大赛冠军等，担任SAT 会议程序委员会主席，受邀到SAT，CP，FMCAD等领域顶级会议做特邀报告。其求解器应用于包括华为、华大九天、中国航空集团、国家电网、阿里巴巴、微软等多家企业/机构在内的多个实际场景，包括集成电路验证、操作系统验证、云计算调度和航空制造调度等。

In the Subset Feedback Arc Set in Tournaments (Subset-FAST) problem, we are given as input a tournament $T$ with a vertex set $V(T)$ and an arc set $A(T)$, along with a terminal set $S \subseteq V(T)$, and an integer $k$. The objective is to determine whether there exists a set $F \subseteq A(T)$ with $|F| \leq k$ such that the resulting graph $T-F$ contains no cycle that includes any vertex of $S$. When $S=V(T)$, this is the classic Feedback Arc Set in Tournaments (FAST) problem. We obtain the first polynomial kernel for this problem parameterized by the solution size. More precisely, we obtain an algorithm that, given an input instance $(T,S,k)$, produces an equivalent instance $(T',S',k')$ with $k' \leq k$ and $|V(T')|=O(k^2)$.

It was known that FAST admits a simple quadratic vertex kernel and a non-trivial linear vertex kernel. However, no such kernel was previously known for Subset-FAST. Our kernel employs variants of the most well-known reduction rules for FAST and introduces two new reduction rules to identify irrelevant vertices. As a result of our kernelization, we also obtain the first sub-exponential time FPT algorithm for Subset-FAST.

Speaker Bio

Tian Bai is currently a postdoctoral researcher at the University of Bergen (UiB). Previously, he worked as a postdoctoral researcher at the University of Hong Kong (HKU) in the ALGO Lab. He received his Ph.D. from the University of Electronic Science and Technology of China (UESTC) as a member of the Algorithms and Logic Group, under the supervision of Prof. Mingyu Xiao.
His research is centered on the design and analysis of algorithms, particularly exact and parameterized algorithms for graph problems. He also has a strong interest in algorithmic game theory.

This talk focuses on the multipacking problem in undirected graphs. We begin by outlining the hardness results established for this problem. A multipacking in an undirected graph $G = (V, E)$ is a set $M \subseteq V$ such that for every vertex $v \in V$ and every integer $r \ge 1$, the ball of radius $r$ around $v$ contains at most $r$ vertices of $M$; equivalently, there are at most $r$ vertices of $M$ within distance $r$ from $v$ in $G$. The multipacking number of $G$ is the maximum cardinality of a multipacking of $G$. The Multipacking problem asks whether a graph contains a multipacking of size at least $k$.

For more than a decade, it remained an open question whether the Multipacking problem is NP-complete or solvable in polynomial time, although the problem is known to be polynomial-time solvable for certain graph classes (such as strongly chordal graphs, grids, etc.). We resolve this open question by proving that the Multipacking problem is NP-complete for undirected graphs. Moreover, we show that the problem is W[2]-hard when parameterized by the solution size. We further strengthen these results by proving that the problem remains NP-complete and W[2]-hard (under the same parameterization) even for several graph subclasses. Additionally, we design an exact exponential-time algorithm that computes a maximum multipacking in an $n$-vertex graph in time $O^*(1.63^n)$.

Finally, we present an approximation algorithm for the Multipacking problem on cactus graph. The talk will outline the proof technique behind the $1.5$-approximation algorithm for cactus graphs.

Speaker Bio

Sk Samim Islam, Ph.D. scholar at the Indian Statistical Institute, Kolkata, India.

Online Access

Join Zoom Meeting
https://us06web.zoom.us/j/87293332992?pwd=RwexmgEeYOaHpBaprbdO4P7FKC7KRV.1

Meeting ID: 872 9333 2992
Passcode: 009710

Modern machine learning models–such as large language models-are increasingly accurate at making predictions, and their outputs are often used by downstream decision-makers to guide actions. For these predictions to be truly useful, however, they must not only be accurate but also reliable. A key criterion for reliability is calibration, which ensures that predicted probabilities align with actual outcomes. In this talk, l present an information-design perspective on calibrated prediction. I will introduce a general framework for comparing the informativeness of different machine predictors, and show how this lens helps us understand the limits and possibilities of calibration. I will then discuss how to optimize predictors in environments where decision-makers incentives may be misaligned with those of the predictor.
Based on the joint work with Wei Tang and Liuhan Qian. Some results have been accepted in SODA 2026。

Speaker Bio

Yiding Feng is an assistant professor at HKUST IEDA. Previously, he worked as a principal researcher atthe University of Chicago Booth School of Business, and postdoctoral researcher at Microsoft Research New England. He received his Ph.D. from the Department of Computer Science, at Northwestern University in 2021. His research focuses on operations research, economics & computation, and theoretical computer science. He was the recipient of the INFORMS Auctions and Market Design (AMD) Michael H. Rothkopf Junior Researcher Paper Prize (second place).

Although spectral sparsification is well understood for undirected graphs, developing efficient algorithms for Eulerian directed graphs — a key step towards efficient directed Laplacian solvers — has long been challenging. Eulerian sparsification requires preserving degree balance while controlling asymmetric matrix variances, making standard undirected techniques insufficient.

In this talk, I will present a line of my recent work that progressively closes this gap. I will describe how effective resistance decomposition and electrical routing enable efficient edge sampling, how matrix discrepancy theory helps reduce sparsity, and how degree-preserving patching graph combined with dynamic expander decomposition yields fully dynamic “Eulerian” sparsifiers. Together, these results bring the theory and efficiency of Eulerian sparsification much closer to that of the undirected setting.

This talk is based on joint works with Arun Jambulapati, Sushant Sachdeva, Aaron Sidford, Anvith Thudi, and Kevin Tian.

Speaker Bio

Yibin Zhao is a final-year PhD candidate at the University of Toronto, advised by Sushant Sachdeva. Previously, he received his B.S. degree in Computer Science from the University of Toronto. His research focuses on the design and analysis of fast algorithms, with particular interest in problems at the intersection of algorithmic graph theory, numerical linear algebra, and data structures.

现有的人工智能技术是一个物理世界科学体系下的、以计算为中心的、数据驱动的智能技术，是一个“有算无谋”的信息处理技术（所谓的算力出奇迹）。这是图灵技术路线的技术，并没有回答智能的实质是什么？智能的原理、策略是什么？等基本问题。人是最高智能体，人工智能就是用机器来实现人的智能。人的智能策略无非就是“谋”和“算”，谋全局，算局部，先谋后算。从计算到谋算是人工智能发展的必由之路。长期以来，一直没有一个谋的数学理论。谋和谋算的数学原理是破解人工智能科学障碍的必要条件。我提出“谋”就是层谱抽象、跨越抽象层谱的演算、推理与优化（直觉推理）的数学定义，“算”就是分而治之，同一抽象层谱的演算、推理与优化的理论。“谋”是信息科学的概念，“算”是物理世界的科学概念。学习、智能、人工智能是信息世界的科学现象，只能在信息世界科学体系下来建立。信息是打开信息世界的钥匙，是建立人工智能科学体系的支点。我建立了信息的数学原理，建立了人工智能的信息科学原理。提出了智能论题：一个主体的智能就是该主体的信息，即

智能=信息。

人工智能的信息科学原理建立了人工智能的信息模型，即孙子模型，奠定了有原理、可解释、高效、可信的人工智能技术、人工智能技术路线和人工智能机器的数学基础。本报告将介绍人工智能的这一科学原理。

Speaker Bio

北京航空航天大学教授。2003年国家杰出青年基金获得者，2008年中国科学院百人计划入选者。现任中国人工智能学会人工智能基础专业委员会主任。

1993年中国科学院软件研究所研究生毕业，获博士学位。1993年7月-2018年7月在中国科学院软件研究所工作，分别于1995，1999年被聘为副研究员、研究员。分别于1998年1月-1999年1月，2000年2月-2002年2月在英国Leeds大学做访问学者、研究员。2008年9月-2009年3月在美国Cornell大学做访问科学家。2012年1月-2012年3月，英国剑桥大学牛顿数学研究所做访问学者。2018年7月-今，入职北京航空航天大学计算机学院。

研究计算、信息与智能的数学原理。2008 – 2016年，提出编码树的概念、结构熵的度量，创立了《结构信息论》。2016 – 2024，提出信息世界的层谱抽象科学范式，建立了包括信息演算理论、信息生成原理和信息解码原理为三大支柱的信息的数学原理；建立了包括（观察）学习的信息理论、自我意识的信息理论和博弈/谋算理论为三大支柱的人工智能信息科学原理。2024年在科学出版社出版《人工智能科学 — 智能的数学原理》和《人工智能原理 — 从计算到谋算的模型、原理与方法》专著。 目前主要研究“谋算智能机器原理”和“《孙子兵法》的人工智能原理”。

The Correlated Pandora’s Problem posed by Chawla et al. (2020) generalizes the classical Pandora’s Problem by allowing the numbers inside the Pandora’s boxes to be correlated. It also generalizes the Min Sum Set Cover problem, and is related to the Uniform Decision Tree problem. This paper gives an optimal 4-approximation for the Correlated Pandora’s Problem, matching the lower bound of 4 from Min Sum Set Cover.

Speaker Bio

Zhiyi Huang is an Associate Professor of Computer Science at the University of Hong Kong. Before joining HKU, he was a postdoc at Stanford University from 2013 to 2014, working with Tim Roughgarden. He earned his Ph.D. from the University of Pennsylvania under the supervision of Sampath Kannan and Aaron Roth in 2013, and a bachelor’s degree in 2008 from the first “Yao Class” founded by Andrew Chi-Chih Yao at Tsinghua University.

Zhiyi works broadly on algorithms, focusing on the role of information—and its flip-side, uncertainty—in computation. He is interested in algorithms for sequential decision-making under uncertainty (online algorithms), learning based on different forms of information (learning theory), incentivizing self-interested agents to share private information (mechanism design), and disclosing one kind of information while keeping the other confidential (differential privacy).

Zhiyi’s research was recognized by several Best Paper Awards, including those from ESA 2024 (Track S), FOCS 2020, and SPAA 2015. He was also the recipient of an Excellent Young Scientists Fund (HK & Macau) by NSFC, an Early Career Award by RGC Hong Kong, a Morris and Dorothy Rubinoff Dissertation Award, and a Simons Graduate Fellowship in Theoretical Computer Science.

We consider the impact of fairness requirements on the social efficiency of truthful mechanisms for trade, focusing on Bayesian bilateral-trade settings. Unlike the full information case in which all gains-from-trade can be realized and equally split between the two parties, in the private information setting, equitability has devastating welfare implications (even if only required to hold ex-ante). We thus search for an alternative fairness notion and suggest requiring the mechanism to be KS-fair: it must ex-ante equalize the fraction of the ideal utilities of the two traders. We show that there is always a KS-fair (simple) truthful mechanism with expected gains-from-trade that are half the optimum, but always ensuring any better fraction is impossible (even when the seller value is zero). We then restrict our attention to trade settings with a zero-value seller and a buyer with value distribution that is Regular or MHR, proving that much better fractions can be obtained under these conditions.

The talk is based on the joint work with Moshe Babaioff and Noam Manaker Morag from the Hebrew University of Jerusalem. The conference version of this paper has been accepted in EC 2025.

Speaker Bio

Yiding Feng is an assistant professor at HKUST IEDA. Previously, he worked as a principal researcher at the University of Chicago Booth School of Business, and postdoctoral researcher at Microsoft Research New England. He received his Ph.D. from the Department of Computer Science, at Northwestern University in 2021. His research focuses on operations research, economics & computation, and theoretical computer science. He was the recipient of the INFORMS Auctions and Market Design (AMD) Michael H. Rothkopf Junior Researcher Paper Prize (second place).

Counting problems concern computing the number of solutions to combinatorial problems or the total weight of such solutions when weights are involved. Counting problems have a long history starting with partition functions in statistical physics introduced in the beginning of 20th century, to foundational work by Valiant, to a long string of results on approximation of counting problems and partition functions by Jerrum, Sinclair and others, to holographic algorithms by Valiant, to complete complexity classifications by the author, Dyer, Richerby, Cai, and Chen. We review the basics of counting, present the major ideas behind some of the approaches, and mention some notable results.

Speaker Bio

Dr Bulatov received his PhD in 1995 from the Ural State University in Ekaterinburg, Russia. His early research area was universal algebra and clone theory. When connections between universal algebra and computer science had been discovered, he pioneered the so-called algebraic approach to the Constraint Satisfaction Problem and related fields. In 2000 Dr Bulatov moved, first, to the University of Oxford, and then to Simon Fraser University in Canada, where he works till now. Dr Bulatov is an expert on applications of algebraic methods in computer science. He is a recipient of the 2021 Godel Prize and 2024 Frontiers of Science Award from the International Congress on Basic sciences.

Poster

The $k$d-tree is one of the most widely used data structures to manage multi-dimensional data. Due to the ever-growing data volume, it is imperative to consider parallelism in $k$d-trees. However, we observed challenges in existing parallel $k$d-tree implementations, for both constructions and updates.

The goal of this work is to develop efficient in-memory $k$d-trees by supporting high parallelism and cache-efficiency. We propose the Pkd-tree (Parallel $k$d-tree), a parallel $k$d-tree that is efficient both in theory
and in practice. The Pkd-tree supports parallel tree construction, batch update (insertion and deletion), and various queries including $k$-nearest neighbor search, range query, and range count. We proved that our
algorithms have strong theoretical bounds in work (sequential time complexity), span (parallelism), and cache complexity. Our key techniques include 1) an efficient construction algorithm that optimizes work, span, and cache complexity simultaneously, and 2) reconstruction-based update algorithms that guarantee the tree to be weight-balanced. With the new algorithmic insights and careful engineering effort, we achieved a highly optimized implementation of the Pkd-tree.

We tested Pkd-tree with various synthetic and real-world datasets, including both uniform and highly skewed data. We compare the Pkd-tree with state-of-the-art parallel $k$d-tree implementations. In all tests, with
better or competitive query performance, Pkd-tree is much faster in construction and updates consistently
than all baselines. We released our code.

Speaker Bio

Ziyang Men is a 3rd-year Ph.D. student at University of California, Riverside, where he is fortune to be supervised by Yihan Sun and Yan Gu. Prior to that, Ziyang received the M.Sc. from University of Copenhagen, and received B.En. from the UESTC.

Ziyang’s research lies on designing parallel spatial partition trees that have good theoretical and practical benefits. He is also interested in topics related to ILP and SAT.

Let $A$ be a class of graphs. Graphs $G,H$ are said to be left (right) $A$-equivalent if for every $F\in A$ the number of homomorphisms from $F$ to $G$ and $H$ (from $G$ and $H$ to $F$) are equal. This line of research has started almost 60 years ago by Lovasz and has received much attention recently. Left equivalence usually corresponds to some structural graph parameters, while right equivalence seem to be expressible in terms of various logics. We introduce the problem, briefly survey recent results and state several future research directions.

Speaker Bio

Dr Bulatov received his PhD in 1995 from the Ural State University in Ekaterinburg, Russia. His early research area was universal algebra and clone theory. When connections between universal algebra and computer science had been discovered, he pioneered the so-called algebraic approach to the Constraint Satisfaction Problem and related fields. In 2000 Dr Bulatov moved, first, to the University of Oxford, and then to Simon Fraser University in Canada, where he works till now. Dr Bulatov is an expert on applications of algebraic methods in computer science. He is a recipient of the 2021 Godel Prize and 2024 Frontiers of Science Award from the International Congress on Basic sciences.

Schedule

• First Lecture:
  • 17:00-18:00 Beijing Time
  • April 18, 2025 (Friday)
• Second Lecture:
  • 10:00-11:00 Beijing Time
  • April 20, 2025 (Sunday)
• Third Lecture:
  • 16:00-17:00 Beijing Time
  • April 23, 2025 (Wednesday)

The broken circuit complex is a generalization of the matroid independence complex. Given that the exchange walk over the bases of a matroid is fast-mixing [Anari et al., 2019], we are interested in whether the exchange walks over the broken circuit complex are also fast-mixing. This may enable us, e.g., to approximately uniformly sample and count the number of acyclic orientations of a graph, which is a #P-hard problem of unknown approximability.

Speaker Bio

Giorgos Mousa obtained his PhD from the University of Edinburgh, where he worked with Professor Heng Guo in the Laboratory for the Foundations of Computer Science. His research interests include matroid theory, Markov chains, and sampling and counting problems.

The active regression problem of the single-index model is to solve $\min_x | f(Ax)-b|_p$, where $A$ is fully accessible and $b$ can only be accessed via entry queries, with the goal of minimizing the number of queries to the entries of $b$. When $f$ is Lipschitz, previous results only obtain constant-factor approximations. I shall present an algorithm that provides a $(1+\epsilon)$-approximation solution by querying $\tilde{O}(d^{\frac{p}{2}\vee 1}/\epsilon^{p\vee 2})$ entries of $b$. I shall also show that this query complexity is optimal up to logarithmic factors for $p\in [1,2]$ and that the $\epsilon$-dependence of $1/\epsilon^p$ is optimal for $p>2$.

Speaker Bio

Yi Li is an associate professor in the Division of Mathematical Sciences and holds a joint appointment in the School of Computing and Data Science at Nanyang Technological University. His main research interests lie in algorithms for massive datasets and sublinear time streaming algorithms, with a particular focus on randomized numerical linear algebra in recent years.

You are going to buy a new apartment where you are allowed to choose a floorplan of the apartment. The development company has an AI expert system which shows you a floorplan. If you do not like the floorplan, it shows you another floorplan. It would be great if they can show you all possible floorplans. It is an enumeration algorithm which can show you all possible floorplans. Given some properties of an object, an enumeration algorithm enumerates all objects with the properties. Enumeration algorithms play a crucial role in the Fourth Industrial Revolution (4IR), which is characterized by the fusion of physical, digital, and biological systems. In big data analysis enumeration algorithms help in systematically listing and analyzing large datasets which is crucial for extracting meaningful insights from big data. Enumeration algorithms are used in various Artificial Intelligence (AI) and Machine Learning (ML) applications to explore different model parameters, features,and structures, helping to enhance the accuracy and efficiency of predictive models.
In the area of robotics and automation enumeration algorithms assist in path planning, task scheduling, and resource allocation for robots and automated systems, enabling them to perform complex tasks with precision. This talk explores the critical role of enumeration algorithms in advancing 4IR technologies, offering a comprehensive overview of their design principles and practical applications. We focus on the theoretical foundations of these algorithms, discussing their impact on problem-solving and decision-making processes. As working examples, we will show efficient algorithms to enumerate all distributions of objects to bins and an algorithm to enumerate all triangulations of a polygon. Attendees will gain insights into designing robust enumeration algorithms tailored to contemporary challenges, equipping them with the tools to harness the full potential of 4IR technologies.

Speaker Bio

Dr. Md. Saidur Rahman, a professor (on deputation) of Bangladesh University of Engineering and Technology (BUET) and a fellow of Bangladesh Academy of Sciences, is currently serving as a member of University Grants Commission of Bangladesh. He is a renowned researcher in the field of graph algorithms and is regarded as an authority on graph drawing algorithms. He has more than 130 publications on algorithms and graph theory in internationally reputed journals and conferences. He has developed many efficient algorithms for finding drawings for planar graphs which have enormous applications in VLSI layout automation, software engineering, DNA recognition, etc. His graduate textbook “Planar Graph Drawing,” written with renowned computer scientist Professor Takao Nishizeki, is considered as the most valuable pioneering work in planar graph drawings. He has co-edited five volumes of LNCS series of Springer and served as guest editors for several reputed journals including Algorithmica and Theoretical Computer science. Currently he is in the editorial board of several international Journals including the Journal of Graph Algorithms and Applications and International Journal of Foundation of Computer Science. His undergraduate textbook “Basic Graph Theory” which has been published by Springer has received around 62,000 downloads within six years. Professor Rahman is leading an enthusiastic research group in Graph Drawing and Information Visualization Laboratory of CSE Department, BUET. He has supervised four Ph. D. theses and 27 M. Sc. Engg. theses. He is a recipient of “BAS Gold Medal 2003” in the junior group, “UGC Award 2004” and the prestigious “Funai Information Technology Award for Young Researchers 2004.” Prof. Rahman is a founder of the International Conference and Workshops on Algorithms and Computation (WALCOM) which has appeared as a very prestigious international venue of theoretical computer science over the years. Professor Rahman worked as a member of expert panels of various national and international bodies. Currently he is a member of Board of Accreditation for Engineering and Technical Education (BAETE), Bangladesh which is a full signatory of Washingtom Accord, and also a member of BAETE Evaluation Team (ET). In BUET administration, he showed excellent administrative capabilities as the Registrar, the Director of institutional Quality Assurance Cell (IQAC) and Head of CSE Department of BUET. He is an ACM Distinguished Speaker and recently named as 2024 ACM Distinguished Member.

Poster

The Parameterized Inapproximability Hypothesis (PIH) asserts that no fixed parameter tractable (FPT) algorithm can distinguish a satisfiable instance of a constraint satisfaction problem (CSP) parameterized by the number of variables, from one where every assignment fails to satisfy 1% of the constraints. PIH plays the role of the PCP theorem in parameterized complexity, with many downstream inapproximability consequences.

This talk will introduce the context and statement of the PIH, and then sketch the ideas behind a recent proof showing that the well-known Exponential Time Hypothesis (ETH) implies the PIH. The approach involves two broad steps. The first identifies a highly structured ETH-hard CSP whose variables take vector values, and whose constraints are either parallel or linear. Both kinds of constraints are then checked with constant soundness via a “parallel PCP of proximity” based on the Hadamard code. We will briefly mention follow-up improvements that employ Reed-Muller code based PCPs to even get near-optimal runtime lower bounds, eg. showing that constant factor approximations to $k$-clique on $n$-vertex graphs requires $n^{k^{1-o(1)}}$ time under ETH.

Based on joint works with Bingkai Lin, Xuandi Ren, Yican Sun, and Kewen Wu.

Speaker Bio

Venkatesan Guruswami is a Professor of Computer Science and Mathematics at UC Berkeley and senior scientist at the Simons Institute for the Theory of Computing. Venkat received his Bachelor’s degree from the Indian Institute of Technology, Madras, and his Ph.D. from MIT.

Venkat’s research interests include coding theory, constraint satisfaction, approximate optimization, and computational complexity. He is the Editor-in-Chief of JACM, and was previously president of the Computational Complexity Foundation. Venkat is a recipient of the Presburger Award, a Simons Investigator award, Guggenheim, Packard and Sloan Fellowships, the ACM Doctoral Dissertation Award, and a distinguished alumnus award from IIT Madras.  He is a fellow of the ACM, IEEE, and AMS.

In a digraph, a dicut is a cut where all the arcs cross in one direction. A dijoin is a subset of arcs that intersects each dicut. Woodall conjectured in 1976 that in every digraph, the minimum size of a dicut equals to the maximum number of disjoint dijoins. However, prior to our work, it was not even known if at least $3$ disjoint dijoins exist in an arbitrary digraph whose minimum dicut size is sufficiently large. By building connections with nowhere-zero (circular) $k$-flows, we prove that every digraph with minimum dicut size $\tau$ contains $\tau/k$ disjoint dijoins if the underlying undirected graph admits a nowhere-zero (circular) $k$-flow. The existence of nowhere-zero $6$-flows in $2$-edge-connected graphs (Seymour 1981) directly leads to the existence of $\tau/6$ disjoint dijoins in any digraph with minimum dicut size tau, which can be found in polynomial time as well. The existence of nowhere-zero circular $(2+1/p)$-flows in $6$p-edge-connected graphs (Lovász et al 2013) directly leads to the existence of $\tau p/(2p+1)$ disjoint dijoins in any digraph with minimum dicut size tau whose underlying undirected graph is $6$p-edge-connected.

Speaker Bio

Siyue Liu is a Ph.D. student in the Algorithms, Combinatorics, and Optimization program at the Tepper School of Business, Carnegie Mellon University. Her research focuses on combinatorial optimization and integer programming. She has published in Mathematical Programming and received the Best Paper Award at IPCO.

The Parameterized Inapproximability Hypothesis (PIH) asserts that no fixed parameter tractable (FPT) algorithm can distinguish a satisfiable CSP instance, parameterized by the number of variables, from one where every assignment fails to satisfy an $\varepsilon$ fraction of constraints for some absolute constant $\varepsilon > 0$. PIH plays the role of the PCP theorem in parameterized complexity. However, PIH has only been established under Gap-ETH, a very strong assumption with an inherent gap.

In this work, we prove PIH under the Exponential Time Hypothesis (ETH). This is the first proof of PIH from a gap-free assumption. Our proof is self-contained and elementary. We identify an ETH-hard CSP whose variables take vector values, and constraints are either linear or of a special parallel structure. Both kinds of constraints can be checked with constant soundness via a “parallel PCP of proximity” based on the Walsh-Hadamard code.

Speaker Bio

林冰凯，南京大学计算机学院教授，博士生导师。博士毕业于日本东京大学，硕士及本科毕业于上海交通大学ACM试点班。研究领域是理论计算机科学，具体领域是参数复杂性(Parameterized Complexity) 及近似算法 (Approximation Algorithms)。主要成果包括独自解决了参数复杂性领域基础性难题——k-BICLIQUE问题的参数复杂性;在图嵌入问题参数复杂性的二分猜想这一长达十多年的公开问题上取得重要进展；对经典NP-难优化问题集合覆盖问题取得首个以及当前最好的参数算法不可近似比；证明了指数时间假设蕴含参数不可近似假设，推动了参数PCP理论的发展。成果获得理论计算机国际一流会议STOC 2024与计算机科学图论领域重要国际会议WG 2017最佳论文奖。两篇单独作者论文分别获国际算法会议SODA 2015最佳论文奖和最佳学生论文奖以及欧洲理论计算机重要会议ICALP 2019最佳论文奖。

Poster

大规模离散优化问题在许多实际应用中扮演着至关重要的角色，涉及物流、交通、资源分配、网络设计等多个领域。这类问题的复杂性源于其决策变量的离散性，通常需要在大量可能的解中寻找最优解。随着数据规模的不断增长和应用需求的日益复杂，传统的求解方法面临着效率和准确性的挑战。本次报告首先介绍求解算法效率的几种策略，包括克服循环搜索、基于变量依赖的预处理过程以及基于问题结构信息的启发式策略。之后，报告将重点围绕两个重要问题的求解：伪布尔优化问题和图同构问题。在伪布尔优化问题的研究中，提出了一种有效的局部搜索框架，显著提升了解的质量和求解的速度。针对图同构问题，提出了一种精确算法，提出了五个关键策略，包括更强的约束传播以简化顶点匹配域，引入结合点度数和解密度的匹配排序方法，以及自适应的约束传播机制。此外，为了更高效地求解大规模实例，提出了增强的边约束方法和域限制策略。实验结果表明，PathLAD+的性能显著优于当前多个最先进的精确算法。

Speaker Bio

王艺源，男，东北师范大学信息科学与技术学院副教授，博士生导师，中国人工智能学会智能服务专委会委员，中国计算机学会理论计算机科学专委会委员，国际期刊IEEE-TLT专刊编委。2017年于吉林大学计算机科学与技术学院获得工学博士学位。2020年获得吉林省青年人才托举计划。主要从事人工智能、算法设计及其应用、逻辑推理、优化求解等方向研究，特别关注大规模组合优化问题求解等。主持并参与多项国家自然基金，在AIJ、JAIR、EJOR、AAAI、IJCAI、CP等上发表学术论文40余篇，其中以第一作者或通讯作者发表计算机学会推荐A类论文10余篇，并多次受邀担任IJCAI、AAAI等会议程序委员会委员。2022年国际MaxSAT比赛中包揽了完备组加权赛道和非加权赛道的冠军和亚军；2024年国际PBO比赛中获得两个赛道第一名，一个赛道第二名以及两个赛道第三名。

Poster

Congestion games are commonly used to model problems in large-scale networks and represent a simple, yet powerful paradigm for selfish resource sharing. These games belong to the larger class of potential games where the pure Nash equilibria is guaranteed to exist. However, the cost of pure Nash equilibria could still be far from minimizing some prescribed measure of social cost. The Price of Anarchy (PoA), which is the largest ratio between the cost of pure Nash equilibria and the cost of the social optimum, is a classic measure of inefficiency in game theory. In this work, we study the PoA of congestion games over special combinatorial structures, specifically series-parallel networks and matroids.

First, we consider symmetric congestion games in series-parallel networks with affine edge delays. For arbitrary networks, Correa et al. (2019) proved a tight upper bound of $5/2$ on the PoA. On the other hand, Fotakis (2010) showed that restricting to the class of extension-parallel networks makes the worst-case PoA decrease to $4/3$. We prove that, for the larger class of series-parallel networks, the PoA is at most $2$, and that it is at least $27/19$ in the worst case, improving both the best-known upper bound and the best-known lower bound.

Next, we consider symmetric congestion games with edge delays that are polynomial functions with highest degree $p$. We construct a family of symmetric congestion games over arbitrary networks which achieves the same worst-case PoA of asymmetric network congestion games given by Aland et al. (2006). We then establish that in games defined over series-parallel networks the PoA cannot exceed $2^{p+1}-1$, which is considerably smaller than the worst-case PoA in arbitrary networks. We also prove that the worst-case PoA, which is sub-linear in extension-parallel networks (Fotakis, 2010), dramatically degrades to exponential in series-parallel networks.

Finally, we study the PoA of congestion games in matroids. We derive new upper bounds on the PoA of symmetric congestion games defined over $k$-uniform matroids and paving matroids with polynomial delay functions of highest degree $p$. Specifically, we show that by restricting the matroid structure to $k$-uniform or paving matroid, there is also a significant improvement of the PoA.

This is a joint work with Dr. Carla Michini.

Speaker Bio

Bainian Hao (郝柏年) has received the Ph.D. degree in Industrial Engineering from University of Wisconsin-Madison, advised by Dr. Carla Michini. He will join the School of Economics and Management at the Chang’an University in Spring 2025.

Bainian is interested in the areas of algorithmic game theory, integer optimization, combinatorial optimization and their applications in the real-world problems. His works has been published in the international journals and conferences such as Mathematical Programming, WINE, and SAGT.

A large proportion of the Bayesian mechanism design literature is restricted to the family of regular distributions [Mye-81] or the family of monotone hazard rate (MHR) distributions [BMP-63], which overshadows this beautiful and well-developed theory. We (re-)introduce two generalizations, the family of quasi-regular distributions and the family of quasi-MHR distributions.
The significance of our new families is manifold. First, their defining conditions are immediate relaxations of the regularity/MHR conditions (i.e., monotonicity of the virtual value functions and/or the hazard rate functions), which reflect economic intuition. Second, they satisfy natural mathematical properties (about order statistics) that are violated by both original families. Third but foremost, numerous results [BK-96, HR-09, CD-15, DRY-15, HR-14, AHNPY-19, JLQTX-19, FLR-19, GHZ-19, JLTX-20, JLX-23, LM-24] established before for regular/MHR distributions now can be generalized, with or even without quantitative losses.
This talk is based on joint work with Yaonan Jin (Huawei TCS Lab).

Speaker Bio

Yiding Feng is an assistant professor at HKUST IEDA. Previously, he worked as a principal researcher at the University of Chicago Booth School of Business, and postdoctoral researcher at Microsoft Research New England. He received his Ph.D. from the Department of Computer Science, at Northwestern University in 2021. His research focuses on operations research, economics & computation, and theoretical computer science. He was the recipient of the INFORMS Auctions and Market Design (AMD) Michael H. Rothkopf Junior Researcher Paper Prize (second place).

Poster

In this lecture we will give an introduction to the theory of computations and proofs. We will start with classical results such as decidable and undecidable problems and the Gödel incompleteness theorems. Then we will go into more advanced topics, discussing what information can be gained from the fact that a function is not only computable but verifiable in some formal system. Finally, we will focus on fast growing functions and examples of combinatorial statements unprovable in Peano arithmetic.

Speaker Bio

Lev D. Beklemishev, 莫斯科国立大学教授，莫斯科斯捷克洛夫数学研究所博士，2019–2022年担任斯捷克洛夫国际数学中心 (SIMC)副主任，2020年至今担任斯捷克洛夫国际数学中心数学逻辑系主任。俄罗斯科学院院士，曾获得莫斯科数学学会奖，2022年俄罗斯科学院“年度最佳论文”。符号逻辑协会(ASL)委员会成员，俄罗斯A.I. Maltsev奖评审。主要研究方向为：数理逻辑，证明论，形式算术，可证性逻辑，模态逻辑。

Related

October 18, 2024

Chengdu A&L Seminars, Seminars

Poster

Streaming algorithms process a stream of data values. Thereby, every datavalue has to be processed immediately on arrival. In other words, the algorithm hasto process its input in one pass. The goal of streaming algorithms is to minimizespace and processing time per data item, Streaming algorithms are often used inapplications where massive amounts of data must be processed. In the lst part of the lectures I will present and analyze three famous streaming algorithms, In the 2nd part of the lectures, I will talk about enumeration algorithms.

Speaker Bio

Professor Markus Lohrey, Head of Theoretical Computer Science Group, The University of Siegen, Germany. His research interests are Decidability and complexity of problems in automata theory and algebra, Combinatorial Group theory, data compression, Logic in computer science and Streaming algorithms. He has published more than 180 papers and 3 books.

Poster

Multiagent systems are a general framework that depicts the interactions between humans and artificial intelligence. Such interactions appear more and more frequently in daily life, making an intense call for the governance of such hybrid systems. In the tradition of human society governance, the notion of responsibility can often serve as a tool to determine the rewards and punishments, while different forms of norms, such as obligation, permission, and prohibition, are often used as instructions. In this talk, the speaker will discuss how the concepts of responsibility and norms can be formally extended to multiagent systems with autonomous agents.

Paper

https://www.ijcai.org/proceedings/2024/394

Online social platforms have become more and more popular, and the dissemination of information on social networks has attracted wide attention of the industries and academia. Aiming at selecting a small subset of nodes with maximum influence on networks, the Influence Maximization (IM) problem has been extensively studied. Since it is #P-hard to compute the influence spread given a seed set, the state-of-art methods, including heuristic and approximation algorithms, faced with great difficulties such as theoretical guarantee, time efficiency, generalization, etc. This makes it unable to adapt to large-scale networks and more complex applications. With the latest achievements of Deep
Reinforcement Learning (DRL) in artificial intelligence and other fields, a lot of works has focused on exploiting DRL to solve the combinatorial optimization problems. Inspired by this, we propose a novel end-to-end DRL framework, ToupleGDD, to address the IM problem which incorporates three coupled graph neural networks for network embedding and double deep Q-networks for parameters learning. Previous efforts to solve the IM problem with DRL trained their models on the subgraph of the whole network, and then tested their performance on the whole graph, which makes the performance of their models unstable among different networks. However, our model is trained on several small randomly generated graphs and tested on completely different networks, and can obtain results that are very close to the state-of-the-art methods. In addition, our model is trained with a small budget, and it can perform well under various large budgets in the test, showing strong generalization ability. Finally, we conduct extensive experiments on synthetic and realistic datasets, and the experimental results prove the effectiveness and superiority of our model.

Speaker Bio

Dr. Weili (Lily) Wu received her MS and PhD degrees in computer science both from University of Minnesota, in 1998 and 2002 respectively. She is currently a full professor and a lab director of the Data Communication and Data Management (DCDM) Laboratory at the Department of Computer Science and Engineering, the University of Texas at Dallas. Her research interest is mainly in Big Data, Social Network, Blockchain Technology, wireless sensor network, IoT, Data Mining. She has published more than 241 journal papers and 103 conference papers in various prestigious journals and conferences such as IEEE/ACM Transactions on Networking, IEEE Trans. Netw. Sci. Eng., Comput. Social Systems, IoT Journal, ACM Transactions on Knowledge Discovery in Data, IEEE Trans. Reliability, IEEE TKDE, Multimedia, ACM Transaction on Sensor Networks (TOSN), IEEE Trans. Netw. Serv. Manag., Wirel. Commun., Mob. Comput., Parallel Distrib. Syst., IEEE ICDCS, INFOCOM, ACM SIGKDD, etc. I’m an associate editor of IJBRA, Computational Social Networks (CSN), SOP Transactions on Wireless Communications (STOWC), DMAA, Journal of
Combinatorial Optimization (JOCO), and Journal of Global Optimization (JOGO).
She is an associate editor of International Journal of Bioinformatics Research and
Applications (IJBRA), Computational Social Networks (CSN), SOP Transactions on Wireless Communications (STOWC), DMAA, Journal of Combinatorial Optimization (JOCO), and Journal of Global Optimization (JOGO).

In this talk, we consider the one-visibility cops and robber game. The one-visibility cops and robber game is a variation of the classic cops and robber game, where one-visibility means that the information of the robber is known to all cops only when the distance between the robber and at least one cop is at most one. We give a lower bound on the one-visibility cop number of general trees. We propose strategies to clear trees according to their structures. We present a linear-time algorithm for computing the one-visibility cop number of trees. We also present an $O(n \log n)$-time algorithm for computing the cop-win strategies of trees, where $n$ is the number of vertices.

Speaker Bio

Boting Yang received the B.Sc. degree in mathematics from Fudan University, the M.Sc. and Ph.D. degrees in mathematics from Xi’an Jiaotong University, and the Ph.D. degree in computer science from Memorial University of Newfoundland. He is currently a professor at the University of Regina. His recent research focuses on Algorithmic Graph Theory, Computational Geometry, and Computational learning Theory.

In the classical survivable network design problem (SNDP), we are given an undirected graph $G=(V,E)$ with costs on edges and a connectivity requirement $k(s,t)$ for each pair of vertices. The goal is to find a minimum-cost subgraph $H\subseteq G$ such that every pair $(s,t)$ are connected by $k(s,t)$ edge or (openly) vertex disjoint paths, abbreviated as EC-SNDP and VC-SNDP, respectively. The seminal result of Jain [FOCS’98, Combinatorica’01] gives a $2$-approximation algorithm for EC-SNDP, and a decade later, an $O(k^3\log n)$-approximation algorithm for VC-SNDP, where $k$ is the largest connectivity requirement, was discovered by Chuzhoy and Khanna [FOCS’09, Theory Comput.’12]. While there is a rich literature on point-to-point settings of SNDP, the viable case of connectivity between subsets is still relatively poorly understood.

This paper concerns the generalization of SNDP into the subset-to-subset setting, namely Group EC-SNDP. We develop the framework, which yields the first non-trivial (true) approximation algorithm for Group EC-SNDP. Previously, only a bicriteria approximation algorithm is known for Group EC-SNDP [Chalermsook, Grandoni, and Laekhanukit, SODA’15], and a true approximation algorithm is known only for the single-source variant with connectivity requirement $k(S,T)\in\{0,1,2\}$ [Gupta, Krishnaswamy, and Ravi, SODA’10; Khandekar, Kortsarz, and Nutov, FSTTCS’09 and Theor. Comput. Sci.’12].

Speaker Bio

Qingyun Chen is a PhD student in EECS at UC Merced, advised by Professor Sungjin Im. His research focuses on theoretical computer science, in particular, approximation algorithms, online algorithms, and hardness of approximations.

Euclid in around  300 BC proved that the sequence of prime numbers is infinite. This sequence starts 2,3,5,7,11, 13, …; the largest currently known prime number  is  obtained by raising 2 to the power of 82589933 and subtracting 1. Each number is a unique product of prime numbers; so the prime numbers  can be seen as the building blocks for all natural numbers. 

The first part of the talk gives an overview of prime numbers, including a fairly recent result of Green and Tao on arithmetic progressions,  and close calls to the Goldbach conjecture due to Wang, Helfgott, and others. 

The second part focusses on computation: how to recognise via an efficient (i.e., polynomial time) algorithm whether a number is prime, and how to use a hypothetical quantum computer to obtain the prime factorisation of a number in polynomial time. The relevance to cryptography will be discussed as well.

Speaker Bio

PhD 1992, Heidelberg

University of Chicago, 1995-2001

University of Auckland, 2002-

2010 ICM speaker in special session

2020 Humboldt research award

The Fisher and Arrow-Debreu markets are the most fundamental models for resource/good allocation and exchange. However, classical markets (i) only consider two types of constraints, i.e., budgets of individual buyers and capacities of goods, and (ii) involve a complete information setting wherein buyers’ utilities are known, and all the transactions happen in a static market. In this talk, we present generalizations of classical market equilibrium price designs to settings when agents arrive at the market sequentially with utilities and budgets/goods that are not known a priori to the market designer. More specifically, we assume that agents arrive sequentially at the market with utility and budget parameters drawn i.i.d. from some distribution and make online or exchanges in real-time. In this setting, we study the performance limitations of static pricing approaches, which set the same prices for all agents; and develop two adaptive pricing algorithms, one with knowledge of the distribution and another that solely relies on past observations of user consumption, i.e., revealed preference feedback, with provable performance guarantees.

Speaker Bio

Yinyu Ye is currently the K.T. Li Professor of Engineering at Department of Management Science and Engineering and Institute of Computational and Mathematical Engineering, Stanford University. His current research topics include Continuous and Discrete Optimization, Data Science and Applications, Numerical Algorithm Design and Analyses, Algorithmic Game/Market Equilibrium。 Operations Research and Management Science etc.; and he was one of the pioneers ofInterior-Point Methods, Conic Linear Programming, Distributionally Robust Optimization, Online Linear Programming and Learning, Algorithm Analyses for Reinforcement Learning and Markov Decision Process, and etc. He has received several scientific awards including, including the 2009 John von Neumann Theory Prize for fundamental sustained contributions to theory in Operations Research and the Management Sciences, the inaugural 2012 ISMP Tseng Lectureship Prize for outstanding contribution to continuous optimization (every three years), the 2014 SIAM Optimization Prize awarded (every three years), etc.. According to Google Scholar, his publications have been cited 58,000 times.

叶荫宇 (Yinyu Ye) 现任斯坦福大学管理科学与工程系及计算数学工程研究院李国鼎讲座教授。他的主要研究方向为连续和离散优化, 数据科学及应用， 数字算法设计及分析， 算法博弈及市场均衡，运筹及管理科学等；他和其他科学家开创了内点优化算法，锥规划模型，分布式鲁棒优化，在线线性规划和学习，强化学习和马可夫过程算法分析等。他多次获得科学奖项: 包括2009约翰·冯·洛伊曼理论奖，国际数学规划2012 Tseng Lectureship Prize（每三年），2014美国应用数学学会优化奖（每三年）等。根据谷歌学术统计，目前他的文章被引用总计超过58,000次。

We discuss how to incorporate data structures based on nested dissection into the interior point methods (IPM) framework to obtain efficient linear programming (LP) solvers for a number of problems, including planar min-cost flow, planar k-multicommodity flow, LPs with low treewidth, and general separable LPs. Based on joint papers with Sally Dong, Yu Gao, Gramoz Goranci, Yin Tat Lee, Lawrence Li, Richard Peng, and Sushant Sachdeva.

Speaker Bio

Guanghao Ye is a PhD student at MIT EECS, advised by Jon Kelner. He previously earned his Bachelor’s and Master’s degrees at the University of Washington, under the supervision of Yin Tat Lee. His research broadly focuses on theoretical computer science, particularly in the algorithmic aspects of optimization and its applications.

会议详情见：https://mp.weixin.qq.com/s/r4w5K4WzHxY3ia1_QuBuzA

Poster

蒙特卡罗法（Monte Carlo methods）是与计算机同一时期诞生的二十世纪最重要的科技产物之一。这一方法利用随机采样来高效计算原本传统确定性方法难以计算的量。它的发现拓展了人类高效计算的边界，深刻地影响了人们对计算本质的理解。在这一计算优越性的帮助下带来的科学新发现也塑造了人们今天对客观世界的认识。

本报告将系统介绍报告人近年来在计算采样理论、以及现代蒙特卡罗算法的设计与分析等方面取得的系统性的重要进展，包括：刻画高位概率分布可高效采样条件的“计算相变”定理；马尔可夫链蒙特卡罗（MCMC）采样的并行与分布式算法和局部与动态算法等。

Speaker Bio

尹一通，南京大学教授，新基石研究员；本科毕业于南京大学，博士毕业于耶鲁大学；博士毕业后在南京大学工作至今，目前担任南京大学理论计算机科学团队负责人。尹一通的研究领域为理论计算机科学，主要研究兴趣包括：随机算法、计算采样，数据结构、并行与分布式计算理论等。在JACM、SICOMP、STOC、FOCS、SODA等理论计算机科学的重要期刊与会议发表论文五十余篇。主持国家重点研发计划项目“数据科学的若干基础理论”，获国家自然科学基金-优秀青年科学基金支持，曾获CCF/IEEE CS青年科学家、CCF优博导师、中创软件人才奖、教育部新世纪人才、南京大学青年五四奖章等荣誉。

The multi-armed bandits (MAB) problem is a classic online learning framework that characterizes the learning process of an agent within an unfamiliar environment. Serving as a basic framework for interactive machine learning, it finds broad practical applications across diverse domains, including recommender systems, online experimentation, game-playing algorithms such as AlphaGo, and combinatorial optimization problems such as SAT solvers. Despite its foundational significance, when the arm space exhibits considerable size, traditional MAB algorithms face the curse of dimensionality. To make the algorithm applicable to large-scale environments and improve learning efficiency, the investigation of side information is widely studied in the literature. In this presentation, I will introduce our research efforts in developing efficient algorithms enriched with graph and feature information and show the corresponding theoretical analysis. Additionally, I will introduce our research progress on tackling general combinatorial action spaces by mining side information on involved arms and multi-agent systems by effectively utilizing information on agents’ observations.

Speaker Bio

Fang Kong is currently a Ph.D. candidate in the John Hopcroft Center for Computer Science, Shanghai Jiao Tong University. She is also a member in Wu Honor Ph.D. Class in Artificial Intelligence. Fang received her B.S. degree from Shandong University in 2020. Her current research interests focus on bandit algorithms and reinforcement learning theory. Her research work has been published in top conferences of machine learning and theoretical computer science such as SODA, COLT (the first from SJTU), ICML and NeurIPS. She also serves as a reviewer for mainstream machine learning conferences. During her Ph.D. studies, Fang has been a research intern or visiting student of the Chinese University of Hong Kong, Tencent, Microsoft Research Asia, and Alibaba Damo Academy.

软硬件系统开发包括不断地进行优化和验证，形式化方法在这个过程中被广泛应用。本讲座结合芯片开发过程中的最常见逻辑表达方式和分析手段，介绍形式化技术。具体包括以下内容：
– 设计的网表表达
– BDD 及其应用
– 逻辑的计算与近似
– 实例：低功耗优化
– 讨论：布尔代数、NP问题、抽象及计算

Speaker Bio

曹皖林，美国Texas A&M 大学博士，合见工业软件集团 fellow。
研究兴趣包括：形式化方法，EDA与FPGA工具开发。

This is a paper from 1994.  (In my opinion this paper is a work of art.)   
The paper proves Robertson’s conjecture on well-quasi-orders by topological minors for classes of graphs that are closed under minors.
 
 Robertson conjectured that classes of graphs are WQO by topological minors 
 if they don’t have arbitrary long double paths. The conjecture has been proved; 
 The first part of the proof was published in 2023.

 I will explain (ordinary) graph minors, well-quasi-orders, topological minors 
 and everything else that is needed in great detail. 
 I expect that I will need at least 3 hours in total for this paper; 
 so probably we will split it up into two Fridays. 

 For my talk I have not prepared any slides; I will use the whiteboard, and 
 put the text of Ding’s paper on the screen. (We will go through it step by step.) 

The plan is that I will speak later about the FPT algorithm by Grohe et al. (2011) for topological minors (if people are interested in that).

This is a paper from 1994.  (In my opinion this paper is a work of art.)   
The paper proves Robertson’s conjecture on well-quasi-orders by topological minors for classes of graphs that are closed under minors.
 
 Robertson conjectured that classes of graphs are WQO by topological minors 
 if they don’t have arbitrary long double paths. The conjecture has been proved; 
 The first part of the proof was published in 2023.

 I will explain (ordinary) graph minors, well-quasi-orders, topological minors 
 and everything else that is needed in great detail. 
 I expect that I will need at least 3 hours in total for this paper; 
 so probably we will split it up into two Fridays. 

 For my talk I have not prepared any slides; I will use the whiteboard, and 
 put the text of Ding’s paper on the screen. (We will go through it step by step.) 

The plan is that I will speak later about the FPT algorithm by Grohe et al. (2011) for topological minors (if people are interested in that).

Treewidth is a useful tool in designing graph algorithms. Although many NP-hard graph problems can be solved in linear time when the input graphs have small treewidth, there are problems which remain hard on graphs of bounded treewidth. In this paper, we consider three vertex selection problems that are W[1]-hard when parameterized by the treewidth of the input graph, namely the capacitated vertex cover problem, the target set selection problem and the vector dominating set problem. We provide two new methods to obtain FPT approximation algorithms for these problems. For the capacitated vertex cover problem and the vector dominating set problem, we obtain $(1+o(1))$-approximation FPT algorithms. For the target set selection problem, we give an FPT algorithm providing a tradeoff between its running time and the approximation ratio.

This talk will cover some new advances in recent parallel algorithm research.  We will introduce a few new parallel algorithms on classic graph problems such as single-source shortest paths (the rho-stepping and the delta*-stepping algorithms), biconnectivity (the FAST-BCC algorithm), and strongly connected components (the BGSS algorithm). These algorithms are simple and have good theoretical guarantees on work, span, and space.  If time permits, we will cover recent advances in computational models for modern architectures, parallel data structures, machine learning applications, etc.

Speaker Bio

Yan Gu is an Assistant Professor in the Computer Science and Engineering (CSE) Department at the University of California, Riverside (UCR), since 2020. Prior to that, he was a postdoc associate at MIT CSAIL in 2019. Before that, he received his Ph.D. degree and Bachelor’s degree from Carnegie Mellon University in 2018 and Tsinghua University in 2012, respectively, both in computer science. His research interest is algorithm design, broadly on both the most canonical problems, and applications in databases, machine learning, graphics, architecture, etc. He has recently received the Best Paper Awards in PPoPP 2023 and ESA 2023, the Best Technical Paper Runner-up at VLDB 2023, and an Outstanding Paper Award from SPAA 2020.

Parallel processors are ubiquitous nowadays and it is almost impossible to find a single-core processor, probably other than a toaster. However, very few courses and online materials cover the basic knowledge for designing parallel algorithms. This talk consists of two parts. The first part will overview the key ideas for parallel computing and programming, such as the computational and programming models for parallel algorithms, and the key concept of the scheduler. The second half of this talk will overview some of the most exciting advancements in parallel algorithm research in recent years, and introduce promising research topics in this scope.

Speaker Bio

Yan Gu is an Assistant Professor in the Computer Science and Engineering (CSE) Department at the University of California, Riverside (UCR), since 2020. Prior to that, he was a postdoc associate at MIT CSAIL in 2019. Before that, he received his Ph.D. degree and Bachelor’s degree from Carnegie Mellon University in 2018 and Tsinghua University in 2012, respectively, both in computer science. His research interest is algorithm design, broadly on both the most canonical problems, and applications in databases, machine learning, graphics, architecture, etc. He has recently received the Best Paper Awards in PPoPP 2023 and ESA 2023, the Best Technical Paper Runner-up at VLDB 2023, and an Outstanding Paper Award from SPAA 2020.

The recent large-scale availability of mobility data, which captures individual mobility patterns, poses novel operational problems that are exciting and challenging. Motivated by this, we introduce and study a variant of the (cost-minimization) facility location problem where each individual is endowed with two locations (hereafter, her home and work locations), and the connection cost is the minimum distance between any of her locations and its closest facility. We design a polynomial-time algorithm whose approximation ratio is at most $3.103$. We complement this positive result by showing that the proposed algorithm is at least a $3.073$-approximation, and there exists no polynomial-time algorithm with an approximation ratio of $2-\epsilon$ under UG-hardness. We further extend our results and analysis to the model where each individual is endowed with $K$ locations. Finally, we conduct numerical experiments over both synthetic data and US census data (for NYC, greater LA, greater DC, and Research Triangle) and evaluate the performance of our algorithms.

Speaker Bio

Yiding Feng is a postdoctoral principal researcher at the University of Chicago Booth School of Business. Previously, he worked as a postdoctoral researcher at Microsoft Research New England from 2021 to 2023. He received his Ph.D. from the Department of Computer Science, Northwestern University in 2021. His primary research focuses on theoretical computer science, economics & computation, and operations research.

Phylogenetic networks are rooted, acyclic directed graph in which leaves are labelled with genes, genomes or species. They are used to model evolution with reticulate events. The reconstruction of phylogenetic networks is an important but challenging problem in phylogenetics and genome evolution, as the space of phylogenetic networks is vast and cannot be sampled well. One approach to the problem is to solve the minimum phylogenetic network problem, in which phylogenetic trees are first inferred, then the smallest phylogenetic network that displays all the trees is computed. A tree-child network is a phylogenetic network satisfying the condition that every non-leaf node has at least one child that is of indegree one. Here, we develop a new method that infers the minimum tree-child network by aligning lineage taxon strings in the phylogenetic trees, where we reduce the tree-child network inference problem to the shortest common supersequence problem. This algorithmic innovation enables us to get around the limitations of the existing programs for phylogenetic network inference. Our new program, named ALTS, is fast enough to infer a tree-child network with a large number of reticulations for a set of up to 100 phylogenetic trees with 50 taxa.

Speaker Bio

Louxin Zhang is currently a Professor at the Department of Mathematics, the National University of Singapore and the Director of the Center for Data Science and Machine Learning. He obtained the BSc in mathematics from Lanzhou University, China, and the PhD degree in computer science from the University of Waterloo, Canada. His current research interests include algorithms in bioinformatics for genomics, complex networks and big data in biomedicine. He has published over 150 refereed journal and conference papers in mathematics, computer science and bioinformatics and is the co-author of a monograph on sequence comparison and a bioinformatics textbook. He is a member of the Editorial Board of Quantitative Biology, Bioinformatics Advances. He also served as the chair (or co-chair) of the program committee the 17th Annual Asia Pacific Bioinformatics Conference, the 18th RECOMB Comparative Genomics, the Evolution & Comparative Genomics track of the 2015 Annual Int’l Conference on Intelligent Systems for Molecular Biology (ISMB) and the European Conference on Computational Biology.

Given a graph $G$, the maximal induced subgraphs problem asks to enumerate all maximal induced subgraphs of $G$ that belong to a certain hereditary graph class. While its optimization version, known as the minimum vertex deletion problem in literature, has been intensively studied, enumeration algorithms were only known for a few simple graph classes, e.g., independent sets, cliques, and forests, until very recently [Conte and Uno, STOC 2019]. There is also a connected variation of this problem, where one is concerned with only those induced subgraphs that are connected. We introduce two new approaches, which enable us to develop algorithms that solve both variations for a number of important graph classes. A general technique that has been proven very powerful in enumeration algorithms is to build a solution map, i.e., a multiple digraph on all the solutions of the problem, and the key of this approach is to make the solution map strongly connected, so that a simple traversal of the solution map solves the problem. First, we introduce retaliation-free paths to certify strong connectedness of the solution map we build. Second, generalizing the idea of Cohen, Kimelfeld, and Sagiv [JCSS 2008], we introduce an apparently very restricted version of the maximal (connected) induced subgraphs problem, and show that it is equivalent to the original problem in terms of solvability in incremental polynomial time. Moreover, we give reductions between the two variations, so that it suffices to solve one of the variations for each class we study. Our work also leads to direct and simpler proofs of several important known results.

Speaker Bio

Dr. Yixin Cao is an Associate Professor of Computing at Hong Kong Polytechnic University. He received the Ph.D. degree in computer science from Texas A&M University, USA in 2012. Before coming back to China, he was a research fellow at Institute for Computer Science and Control, Hungarian Academy of Sciences. His research interests are in algorithmic graph theory, fine-grained complexity and algorithm design, combinatorial optimization, and their usages in bioinformatics and social networks. His research is supported by the Hong Kong Research Grants Council (RGC) and the National Natural Science Foundation of China (NSFC).

In this paper we consider a defending problem on a network. In the model, the defender holds a total defending resource of R, which can be distributed to the nodes of the network. The defending resource allocated to a node can be shared by its neighbors. There is a weight associated with every edge that represents the efficiency defending resources are shared between neighboring nodes. We consider the setting when each attack can affect not only the target node, but its neighbors as well. Assuming that nodes in the network have different treasures to defend and different defending requirements, the defender aims at allocating the defending resource to the nodes to minimize the loss due to attack. We give polynomial time exact algorithms for two important special cases of the network defending problem. For the case when an attack can only affect the target node, we present an LP-based exact algorithm. For the case when defending resources cannot be shared, we present a max-flow-based exact algorithm. We show that the general problem is NP-hard, and we give a 2-approximation algorithm based on LP-rounding. Moreover, by giving a matching lower bound of 2 on the integrality gap on the LP relaxation, we show that our rounding is tight.

Speaker Bio

Minming Li is currently a professor in Department of Computer Science, City University of Hong Kong. He received his Ph. D. and B.E. degree in the Department of Computer Science and Technology at Tsinghua University in 2006 and 2002 respectively. His research interests include algorithmic game theory, combinatorial optimization and  algorithm design and analysis for scheduling problems.

As a means to facilitate efficient resource allocation, auctions are a fundamental tool in the modern economy and play a pivotal role in mechanism design theory. This talk will provide a brief introduction to auction design. We will discuss fundamental concepts, as well as some most recent advances in algorithmic auction design. We will also explore one of the most successful stories of auction theory being applied to modern online markets, the sponsored search auctions, and present the challenges therein from a computational viewpoint.

Speaker Bio

贝小辉，新加坡南洋理工大学数学系副教授。博士毕业于清华大学交叉信息研究院。他的研究方向为计算经济学以及理论计算机。他曾在STOC、SODA、EC等会议上发表论文40余篇。曾获微软奖学金，南洋助理教授奖，并指导学生获AAAI-2020最佳学生论文奖。

平面图上计数问题的计算复杂性分类探索特定框架下的计数问题是否分为如下的三类：（1）一般图上可解的问题；（2）一般图上#P-难，但在平面图上可解的问题；（3）在平面图上#P-难的问题。其中一个重要的问题是基于匹配门的全息算法对（2）中的问题是否具有通用性。本报告将介绍平面图上计数问题计算复杂性分类的进展，尤其将介绍基于匹配门的全息算法与（2）中问题的关系。

Speaker Bio

付治国，东北师范大学信息科学与技术学院教授，博士生导师，副院长。博士毕业于吉林大学数学学院计算数学专业，美国威斯康星大学麦迪逊分校博士后。研究领域为计数问题的算法与计算复杂性，成果发表在STOC，FOCS， SODA，Information and Computation等理论计算机顶级会议和期刊。

Signed graphs model complex relations using both positive and negative edges. Signed graph neural networks (SGNN) are powerful tools to analyze signed graphs. We address the vulnerability of SGNN to potential edge noise in the input graph. Our goal is to strengthen existing SGNN allowing them to withstand edge noises by extracting robust representations for signed graphs. First, we analyze the expressiveness of SGNN using an extended Weisfeiler- Lehman (WL) graph isomorphism test and identify the limitations to SGNN over triangles that are unbalanced. Then, we design some structure-based regularizers to be used in conjunction with an SGNN that highlight intrinsic properties of a signed graph. The tools and insights above allow us to propose a novel framework, Robust Signed Graph Neural Network (RSGNN), which adopts a dual architecture that simultaneously denoises the graph while learning node representations. We validate the performance of our model empirically on four real-world signed graph datasets, i.e., Bitcoin_OTC, Bitcoin_Alpha, Epinion and Slashdot, RSGNN can clearly improve the robustness of popular SGNN models. When the signed graphs are affected by random noise, our method outperforms baselines by up to 9.35% Binary-F1 for link sign prediction.

Speaker Bio

张泽宇，奥克兰大学计算机科学系在读博士。主要研究方向是社交网络分析，图神经网络。在AAAI、WWW、Sigir等会议上发表多篇学术论文。

产业界（工厂、仓储、港口、矿山、物流等）面临大量复杂的运筹优化问题，如何高效地解决这些问题，是相关企业的核心能力之一。然而，实际业务一般非常复杂，而且随着时间迅速变化，数据收集也往往相当困难，且存在误差。因此，学术界常用的优化方法往往难以直接应用于产业界。如何跨越学术研究与产业应用之间的鸿沟，成为许多运筹优化研究者所面临的问题。

Speaker Bio

付樟华，本硕博均毕业于华中科技大学，2012年至2015年留学法国从事博士后研究。2015年回国后担任香港中文大学（深圳）以及深圳市人工智能与机器人研究院研究员，主要从事人工智能与机器人相关研究。曾于2014年12月夺得运筹优化领域著名的国际竞赛-第11届DIMACS国际算法设计大赛的冠军（华人首冠）。2016年入选深圳市“孔雀计划”海外高层次人才。2022年参与华为珠峰计划，帮助华为解决实际业务中的核心难题，获得华为颁发的“难题火花奖”。目前专注于机器人及人工智能领域相关产业化工作，已服务近十家行业龙头企业。