学术报告:Incremental Linear Discriminant Analysis(LDA) for Data Dimensionality Reduction

报告题目: Incremental Linear Discriminant Analysis(LDA) for Data Dimensionality Reduction

报告人: Prof. Delin Chu (新加坡国立大学)

时间: 201366日(周四)下午3:30-4:30

地点: 主楼409

摘要: It has been a challenge problem to develop fast and efficient incremental linear discriminant analysis (LDA) algorithms although several incremental LDA algorithms have been proposed in the past. For this purpose, we conduct a new study on LDA in this paper and develop a new and efficient incremental LDA algorithm. We first propose a new batch LDA algorithm called LDA/QR which only depends on the data matrix and the sizes of data classes. LDA/QR is obtained by computing the economic QR factorization of the data matrix followed by solving a lower triangular linear system. Hence, LDA/QR is a simple and fast LDA algorithm. The relationship between LDA/QR and Uncorrelated LDA (ULDA) is also revealed. Based on LDA/QR, we develop a new incremental LDA algorithm called ILDA/QR which is the exact incremental version of LDA/QR. The main features of our incremental LDA algorithm ILDA/QR include: (i) it can easily handle not only the case that only one new sample is inserted but also the case that a chunk of new samples are added; (ii) it has pleasant computational complexity and space complexity; and (iii) it is very fast and always achieves comparative classification accuracy compared with ULDA algorithm and existing incremental LDA algorithms. Numerical experiments using some real world data demonstrate that our ILDA/QR is very efficient and competitive with the state-of-the-art incremental LDA algorithms in terms of classification accuracy, computational complexity and space complexity.

讲座通知——Web spam detection using machine learning techniques

澳门科技大学资讯科技学院院长蔡亚从教授(Prof. TSOI Ah-Chung)于6月3日来北邮作学术报告。欢迎感兴趣的同学老师踊跃参与。

讲座题目:Web spam detection using machine learning techniques

主讲人:TSOI Ah-Chung教授(澳门科技大学)



报告摘要:Web spam detection is a challenging problem primarily because of the large number of features involved, the scarcity of having the number of validated spam sites, and that the data is related to one another through a web topology. In this talk, we will use machine learning techniques, e.g., self organising map, and multilayer perceptrons, except that in both cases, they are extended to handle graph data inputs, balancing of imbalanced data, feature reduction, to study the problem. We applied our techniques to two well known publicly available web spam detection datasets, namely, the UK 2006 dataset, and the UK 2007 dataset. On both datasets using our methods, we achieve the best generalisation results so far, published by others. Moreover our method can be applied to other graph based datasets, without much changes, e.g., the mutagenesis dataset, and we also achieve the best results so far. Hence, our techniques is a general methodology which can handle long term dependency in deep learning architectures.


NIST TAC中KBP评测负责人季姮教授2013年6月7日访问实验室

NIST TAC中KBP评测负责人季姮教授将于2013年6月7日下午访问实验室,地点在718会议室,请感兴趣的同学参加。


Heng Ji is an assistant professor in Computer Science at Queens College, and a doctoral faculty member in the Computer Science Department and Linguistics Department at the Graduate Center of City University of New York. She received her Ph.D. in Computer Science from New York University in 2007. Her research interests focus on Natural Language Processing, especially on Cross-source Information Extraction and Knowledge Base Population. She has published over 90 papers. Her recent work on uncertainty reduction for Information Extraction was invited for publication in the Centennial Year Celebration of IEEE Proceedings. She received a Google Research Award in 2009, NSF CAREER award in 2010, Sloan Junior Faculty award and IBM Watson Faculty award in 2012. She served as the coordinator of the NIST TAC Knowledge Base Population task in 2010 and 2011, the Information Extraction area chair of NAACL-HLT2012 and ACL2013 and the co-leader of the information fusion task of ARL NS-CTA program in 2011 and 2012. Her research has been funded by NSF, ARL, DARPA, Google and IBM.


christian louboutin sale,ralph lauren outlet online,louis vuitton sale,michael kors bags outlet,wedding dresses online australia

美国雪城大学(Syracuse University)唐剑博士访问实验室

报告题目:Sensing-as-a-Service: A Cloud Computing System for Mobile Phone Sensing

报告人:美国雪城大学(Syracuse University) 唐剑博士

主持人:北邮模式识别实验室 肖波副教授

时间:  12月13日(周四)下午15:00 – 16:30

地点:   教3楼8层811会议室

Dr. Jian TANG

Abstract: Sensors on (or attached to) mobile phones can enable attractive sensing applications in different domains such as environmental monitoring, social networking, healthcare, etc. In this talk, I will introduce a new concept, Sensing-as-a-Service, i.e., providing sensing services using mobile phones via a cloud computing system. A Sensing-as-a-Service cloud should meet the following requirements: 1)It must be able to support various mobile phone sensing applications on different smartphone platforms. 2) It must be energy-efficient. 3) It must have effective incentive mechanisms that can be used to attract mobile users to participate in sensing activities. I will identify unique challenges of designing and implementing a Sensing-as-a-Service cloud, review existing systems and methods, present viable solutions, and point out future research directions.

Speaker’s Bio:  Dr. Jian Tang is currently an Assistant Professor in the Department of Electrical Engineering and Computer Science at Syracuse University. He earned his Ph.D degree in Computer Science from Arizona State University in 2006. His research interests lie in the areas of Cloud Computing, Big Data and Wireless Networking. Dr. Tang has published over 60 papers in premier journals and conferences. He received an NSF CAREER award in 2009. He has been an associate editor of IEEE Transactions on Vehicular Technology  since 2010. He served as a co-chair for the Wireless Networking Symposium of IEEE Globecom’2010, ICNC’2012 and ICNC’2013. He also served as a TPC member for many conferences including IEEE Infocom 2010-2013, ICC 2006-2012 and Globecom 2006-2012.



1、       TAC Knowledge Base Population (KBP) 2012


KBP评测的目标是促进自动化系统的研究,包括在大语料库中发现命名实体信息,以及将这些信息整合到知识库中。TAC 2012的任务有三块,均旨在提高从文本中自动填充知识库的能力:

Entity-Linking:给定一个query包含一个名字字符串,一个背景文档ID,一组标识字符串起始位置的UTF-8码,系统需要输出相应名字指示的KB entry的ID号,如果没有,输出一个”NILxxxx” ID。

Slot-Filling:给定一个命名实体、一个预定义的属性集,通过抓取相关值的信息,扩充一个KB节点的属性值。KB的参考文献来自英文维基。Slot Filler Validation这个诊断任务,将去判断参赛者的系统是否正确完成了填充。

Cold Start Knowledge Base Population:给定一个KB概要(空的知识库),通过挖掘大文本数据来构建出KB。



图1.KBP 2012 Slot-Filling前6名(PRIS名列第一)


图2. KBP 2012 Slot-Filling team(按年度,名次变化)


伊利诺伊大学芝加哥分校、都柏林大学、德国萨尔大学、纽约大学、NEC Laboratories、广东外国语大学、香港理工大学、北京邮电大学、中国科学院等。




学术报告通知: Bayesian Inference with Max-margin Posterior Regularization

报告题目: Bayesian Inference with Max-margin Posterior Regularization
报告人:朱军 副教授 清华大学智能技术与系统国家重点实验室智能媒体组
时间:2012年11月23日 10:00-11:30
报告摘要:Existing Bayesian models, especially nonparametric Bayesian methods, rely heavily on specially conceived priors to incorporate domain knowledge for discovering improved latent representations. While priors can affect posterior distributions through Bayes’ theorem, imposing posterior regularization is arguably more direct and in some cases can be more natural and easier. In this talk, I will present regularized Bayesian inference (RegBayes), a computational framework to perform posterior inference with a convex regularization on the desired post-data posterior distributions. When the convex regularization is induced from a linear operator on the posterior distributions, RegBayes can be solved with convex analysis theory. Furthermore, I will present some concrete examples, including MedLDA for learning discriminative topic representations and infinite latent support vector machines for learning discriminative latent features for classification. All these models explore the large-margin idea in combination with a (nonparametric) Bayesian model for discovering predictive latent representations. I will discuss both variational and Monte Carlo methods for approximate inference.
报告人简介:Dr. Jun Zhu is an associate professor in the Department of Computer Science and Technology at Tsinghua University. His principal research interests lie in the development of statistical machine learning methods for solving scientific and engineering problems arising from artificial and biological learning, reasoning, and decision-making in the high-dimensional and dynamic worlds. Prof. Zhu received his Ph.D. in Computer Science from Tsinghua University, and his advisor was Prof. Bo Zhang. He did post-doctoral research with Prof. Eric P. Xing in the Machine Learning Department at Carnegie Mellon University. His current work involves both the foundations of statistical learning, including theory and algorithms for probabilistic latent variable models, sparse learning in high dimensions, Bayesian nonparametrics, and large-margin learning; and the application of statistical learning in social network analysis, data mining, and multi-media data analysis.

伦敦大学玛丽女王学院Shaogang Gong教授访问实验室

9月24日,应实验室郭军老师的邀请,英国伦敦大学玛丽女王学院Shaogang Gong教授访问实验室,并受聘为我校高等智能与网络服务创新引智基地(111基地)长期高级访问科学家,授予仪式在行政办公楼501会议室举行。我校高等智能与网络服务创新引智基地主任、副校长郭军教授出席了授予仪式并在授予仪式前亲切会见了Gong教授及陪同访问的英国伦敦大学玛丽女王学院Tao Xiang博士和Yizhe Song博士。国际处任春霞处长,研究生院副院长、国际学院副院长张琳教授,高等智能与网络服务创新引智基地秘书王小捷教授,信息与通信工程学院杨洁副院长、及实验室的部分师生参加了授予仪式。

授予仪式由高等智能与网络服务创新引智基地王小捷教授主持。郭军副校长代表高等智能与网络服务创新引智基地致词,对Gong教授受聘为我校高等智能与网络服务创新引智基地长期高级访问科学家表示祝贺,并希望双方未来能够有一个良好的合作前景。张洪刚老师介绍了Gong教授的履历。随后,郭军副校长为Gong教授颁发聘书并佩戴校徽。Gong教授对北京邮电大学给予他这个殊荣表示了由衷的感谢。接下来,Gong教授作了题为“From Surveillance to Big Data”的专题学术报告,受到参会师生的热烈欢迎。

2(6) 1(13) 0(13)


报告题目:Image Recognition via Feature Learning

报告人:百度技术副总监、多媒体部负责人  余凯 博士

主持人:北邮模式识别实验室 高升

时间:  11月14日(周三)下午14:30 – 16:30

地点:   教3楼8层811会议室

Dr.  Kai YU

Abstract: The quality of visual features is crucial for a wide range of computer vision topics, e.g., scene classification, object recognition, and object detection, which are very popular in recent computer vision venues. All these image classification tasks have traditionally relied on hand-crafted features to try to capture the essence of different visual patterns. Fundamentally, a long-term goal in AI research is to build intelligent systems that can automatically learn meaningful feature representations from a massive amount of image data.

The primary objective of this talk is to introduce a paradigm of feature learning from unlabeled images, with an emphasis on applications to supervised image classification. We provide a comprehensive coverage of recently developed algorithms for learning powerful sparse nonlinear features, and showcase their superior performance on a number of challenging image classification benchmarks, including Caltech101, PASCAL, and the recent large-scale problem ImageNet. Furthermore, we describe deep learning and a variety of deep learning algorithms, which learn rich feature hierarchies from unlabeled data and can capture complex invariance in visual patterns.


余凯博士任百度技术副总监,多媒体部负责人,主要负责公司在语音、图像、音频等领域面向互联网和移动应用的技术研发。加盟百度前,余凯博士在美国NEC研究院担任Media Analytics部门主管(Department Head),领导团队在机器学习、图像识别、多媒体检索、视频监控,以及数据挖掘和人机交互等方面的产品技术研发。此前他曾在西门子公司任Senior Research Scientist。2011年曾在斯坦福大学计算机系客座主讲课程“CS121: 人工智能概论”。他在NIPS, ICML, CVPR, ICCV, ECCV,SIGIR, SIGKDD,TPAMI,TKDE等会议和杂志上发表了70多篇论文,引用超过2800次,H-index=28。曾担任机器学习国际会议ICML10, ICML11, NIPS11, NIPS12的Area Chair. 2012年他被评为中关村高端领军人才和北京市海聚计划高层次海外人才。余凯博士在南京大学获得学士和硕士学位,在德国慕尼黑大学获博士学位。