论坛简介
知识引擎是人工智能的核心和基础设施。 大数据环境下数据的分布、异构、动态、碎片化和低质等特征给知识工程和知识服务提出了新挑战，既需要从感知角度学习数据的分布表示，又需要从认知角度解释数据的语义，构建新一代知识图谱，暨认知图谱成为实现人工智能的关键。本论坛将邀请认知计算与知识工程领域的著名研究者，一起探讨如何通过半自动化的交互学习构建超大规模认知图谱，并探讨其在实际系统中的可能应用，将深入讨论机器学习和符号计算相结合的知识表示、获取、推理和知识服务理论和方法。
主题报告
Huan Liu, 教授，亚利桑那州立大学
题目：The need for knowledge in Machine learning in the age of big data
摘要：Big data is pervasive and getting big. The unparalleled computing power combined with big data has shown unprecedented success in many applications. In the awe and praise of the unexpected effectiveness of big data, we seem to forget about the role of knowledge in machine learning, which played an important role in the early days of AI. In this talk, we look at the need for knowledge in our endeavor to go beyond the state-of-art machine learning through case studies and hope to pique the curiosity of using knowledge in machine learning as an external source to constrain the hypothesis space and as a means to reuse the rich results of machine learning in many successful applications.
个人简介：Dr. Huan Liu is a professor of Computer Science and Engineering at Arizona State University. His research interests are in data mining, machine learning, social computing, and artificial intelligence, investigating problems that arise in real-world applications with high-dimensional data of disparate forms. He co-authored the textbook on Social Media Mining: An Introduction by Cambridge University Press. He is a founding organizer of the International Conference Series on Social Computing, Behavioral-Cultural Modeling, and Prediction, and Chief Editor of Data Mining and Management in Frontiers in Big Data. He is a Fellow of ACM, AAAI, AAAS, and IEEE.
杨红霞，资深算法专家，阿里巴巴达摩院
题目： Graph Neural Network in Practice
摘要：超大规模图计算在大数据公司发挥着越来越重要的作用，尤其是以深度学习和图计算结合的大规模图表征为代表的系列算法在阿里的诸多业务场景都取得了阶段性成果。阿里巴巴生态体系的数据是极其丰富多样的，涵盖着从购物、出行到娱乐、支付等各个方面信息。比如个性化推荐系统作为一种信息过滤的重要手段，可以依据我们的习惯和爱好推荐合适的商品或者服务，但是传统的推荐系统容易出现稀疏性、冷启动和信息重复性的问题，而认知图谱作为大规模图表征的一类重要模型正在成为一种新兴类型的推荐辅助信息，可以将自然人的全域信息有效的结合起来，真正挖掘、理解每个人的需求，在每个业务域推出消费者真正感兴趣的潜在商品。我们正在致力于研发新一代基于超大规模认知图谱/图计算的推理引擎，可以高效的对于超过几十亿节点，千亿的边进行推理分析。基于这个挑战，我们同时也在信息检索、分布式计算、大规模系统设计、机器学习、人工智能、自然语言处理等相关领域带来新鲜理念和突破性进展。
个人简介：美国杜克大学博士学位，原IBM全球研发中心Watson研究员，Yahoo！主任数据科学家，现任数据技术发展部资深算法专家，带领团队开发基于数据中台的智能算法，稳定的支持了阿里巴巴搜索、广告等30几个核心BU和其业务场景。在顶级统计和机器学习国际学术期刊会议发表论文28篇（包括JASA, ICML, ATSTATS, KDD, ICDM，CIKM和WWW等），美国专利9项，任职Applied Stochastic Models in Business and Industry副主编，International Statistical Institue理事。
王冠三，高级研究经理，Microsoft Research
题目：学术图谱中的认知问题
摘要：介绍微软学术搜索中知识图谱的构建方法和技巧，并探讨其中碰到的认知问题。
个人简介：Kuansan Wang is a Managing Director of Microsoft Research Outreach, responsible for Microsoft Academic Services. He is also a principal researcher of MSR Redmond, conducting research in web search, data mining, natural language processing and information retrieval. He joined MSR Speech technology group in 1998, working on acoustic modeling, language modeling and spoken and multimodal systems. He has contributed to numerous Microsoft speech products as well as many international standards in the speech areas. He started working on web search in 2007 and became one of the earliest members in the newly created Internet Service Research Center (ISRC). Dr. Wang received his PhD and MS from University of Maryland, College Park, and BS from National Taiwan University, all in Electrical Engineering.
金国庆，教授，香港中文大学
题目：社交网络分析中的图嵌入：最新进展和未来方向
摘要：社交网络分析描述社交结构（例如社交媒体网络等）内部的社交互动，是一个至关重要且具有挑战性的研究课题。图嵌入是指保留图的结构信息的同时，将图中的元素（包括节点，边，子结构或整个图）投影到低维空间的问题。图嵌入是分析大规模社交网络的一项重要技术。为了应对现代社交媒体网络的规模不断扩大和结构多样化的问题，研究人员提出了新的图嵌入方法。在本次演讲中，我们将介绍图嵌入的最新进展和未来方向，以及在社交网络分析中的常见应用，例如节点分类， 链接预测，社区检测和社交推荐。我们将介绍传统图嵌入方法（例如矩阵分解）和最近流行的基于深度学习的图嵌入方法。最后，我们将指出一些有潜力的研究方向作为未来的工作。
个人简介：金教授的研究兴趣包括机器学习，社会计算，人工智能，网络智能，数据挖掘和多媒体信息处理。在这些研究领域，他发表顶级期刊和会议论文累计300余篇。他现任ACM Transactions on Knowledge Discovery from Data (ACM TKDD) 副主编和 Journal of Neural Network副主编。他是国际神经网络协会（INNS）主席和IEEE院士。此外，他还是WebConf 2020, ICONIP2020, WSDM2011, RecSys2013, ACML2015的联合主席，并且在各个顶级会议 （例如WWW, NIPS, ICML, IJCAI, AAAI等）担任多种职务。他是香港中文大学计算机科学与工程系教授并曾任工程学院副院长（教育）。他曾休假于旧金山的AT&T实验室从事科研工作并于加州大学伯克利分校讲授专业课程《社会计算与数据挖掘》。他于加州理工学院（帕萨迪纳）获得工程与应用科学学士学位，于南加州大学洛杉矶分校获得计算机科学的硕士和博士学位。
李朝，资深数据科学家，阿里巴巴
题目：基于大规模图的电商层次主题分类
摘要：E-commerce taxonomy plays an essential role in online retail business. Existing taxonomy of e-commerce platforms organizes items into an ontology structure. However, the ontology-driven approach is subject to costly manual maintenance and often does not capture user's search intention, particularly when user searches by her personalized needs rather than a universal definition of the items. Observing that search queries can effectively express user's intention, we present a novel large-Scale Hierarchical taxOnomy via grAph based query coaLition (SHOAL) to bridge the gap between item taxonomy and user search intention. SHOAL organizes hundreds of millions of items into a hierarchical topic structure. Each topic that consists of a cluster of items denotes a conceptual shopping scenario, and is tagged with easy-to-interpret descriptions extracted from search queries. Furthermore, SHOAL establishes correlation between categories of ontology-driven taxonomy, and offers opportunities for explainable recommendation. The feedback from domain experts shows that SHOAL achieves a precision of 98% in terms of placing items into the right topics, and the result of an online A/B test demonstrates that SHOAL boosts the Click Through Rate (CTR) by 5%. SHOAL has been deployed in Alibaba and supports millions of searches for on-line shopping per day.
个人简介：Dr. Zhao Li is currently a senior staff data scientist at Alibaba Group, specializing in adversarial intelligence with attention to e-commence ranking and recommendation systems. He previously worked at TCL Research America as a chief data scientist on recommendation systems. He received a fellowship from NSF and completed a Ph.D. with graduate award in Computer Science Department from University of Vermont. Prior to that, he obtained his M.S. from South China University of Technology. He has published over 50 papers in prestigious conferences including NIPS, IJCAI, AAAI, KDD, etc,. His current research interests include Adversarial Machine Learning, Network Representation Learning, Multi-Agent Reinforcement Learning, and Big Data Driven Security. He is a Technical Committee member of CCF on database. Has has served on program committees of conferences, including AAAI, AAMAS, CIKM, and ICME.
唐杰，教授，清华大学
摘要：We propose a novel CognitiveGraph framework for learning with knowledge graphs. Inferring new facts from existing knowledge graphs (KG) with explainable reasoning processes is a significant problem and has received much attention recently. The proposed CognitiveGraph framework is founded on the dual process theory in cognitive science. The framework gradually builds a cognitive graph in an iterative process by coordinating an implicit extraction module (System 1) and an explicit reasoning module (System 2). While giving accurate answers, our framework further provides explainable reasoning paths. Specifically, our implementation based on BERT and graph neural network (GNN) efficiently handles graph with tens of millions of nodes. The framework has many applications. For example, for multi-hop reasoning-based QA (e.g., HotpotQA), it achieves a winning joint F1 score of 34.9 on the leaderboard, compared to 23.6 of the best competitor.
个人简介：清华大学计算机系长聘教授、计算机系副主任、清华-工程院知识智能联合实验室主任，获杰青。研究兴趣包括：社会网络分析、数据挖掘、机器学习和知识图谱。发表论文200余篇，引用10000多次（个人h-指数59）。主持研发了研究者社会网络挖掘系统AMiner，收录1.36亿科研人员、2.31亿科技文献，吸引了220个国家/地区1000多万独立IP访问。曾担任国际期刊ACM TKDD的执行主编和国际会议CIKM’16、WSDM’15的程序委员会主席、KDD’18大会副主席以及IEEE TKDE、ACM TIST、IEEE TBD等期刊编委。作为第1完成人获北京市科技进步一等奖、中国人工智能学会科技进步一等奖、KDD杰出贡献奖。
主席
往期回顾
CogGraph'18, @ CNCC2018 • 杭州 • 2018年10月25日