Abstract

Internet of Things (IoT) connects everyday objects and changing environments, to the Internet to enable communication/interactions between these objects and people. IoT is accelerating the growth of data available on the Internet, which makes the traditional search paradigms incapable of digging the information that people need from massive and deep resources. Furthermore, given the dynamic nature of organizations, social structures, and devices involved in IoT environments, intelligent and automated approaches become critical to support decision makers with the knowledge derived from the vast amount of information available through IoT networks. This talk will share some of our research on building an effective and efficient paradigm of “proactive discovering” Things-of-Interest (TOI) and provides an array of new perspectives on enabling an interactive Internet of Things to unleash the full power of IoT.

BIO

A/Prof. Lina Yao is currently a Scientia Associate Professor at University of New South Wales (UNSW), Australia. Her research lies in data mining and machine learning with focus on recommender systems, activity recognition, Brain Computer Interface and Internet of Things. She has published over 150 peer-reviewed papers in prestigious journals and top international conferences in the areas of data mining, machine learning and intelligent systems including ACM CACM, ACM CUSR, IEEE TMC, ACM TIST, IEEE TKDE, ACM TKDD, ACM TOIT, PR, IEEE TNSRE, IEEE TNNLS, IEEE CYB, IEEE IIT, JBHI, NeurIPS, SIGKDD, ICDM, UbiComp, AAAI, SIGIR, IJCAI and CIKM. She is serving as the Associate Editor for ACM Transactions on Sensor Networks (TOSN).

Abstract

With the rapid population growth and urbanization, smart cities and urban computing are emerging as the priority for research and development across the world. Big human mobility and urban sensing data are increasingly produced by the sensors ofInternet of Things (IoT) via emerging communication technologies. The effective use of these big data can certainly help create smart cities where infrastructure and resources are used in a more efficient manner. Thus, my research aims to enable urban intelligence with big data, and develop the novel algorithms, cutting-edge technologies, and applicable systems for next-generation urban management. In this talk, I will briefly introduce my several research projects on these topics, such as big data for next-generation emergency response and disaster management and UrbanBrain for next-generation urban management.

BIO

Prof. Xuan Song received the Ph.D. degree in signal and information processing from Peking University in 2010. In 2017, he was selected as Excellent Young Researcher of Japan MEXT. In the past ten years, he led and participated in many important projects as principal investigator or primary actor in Japan, such as DIAS/GRENE Grant of MEXT, Japan; Japan/US Big Data and Disaster Project of JST, Japan; Young Scientists Grant and Scientific Research Grant of MEXT, Japan; Research Grant of MLIT, Japan; CORE Project of Microsoft; Grant of JR EAST Company and Hitachi Company, Japan. He served as Associate Editor, Guest Editor,Program Chair, Area Chair, Program Committee Member or reviewer for many famous journals and top-tier conferences, such as IMWUT, IEEE Transactions on Multimedia, WWW Journal, Big Data Journal, ISTC, MIPR, ACM TIST, IEEE TKDE, UbiComp, ICCV, CVPR, ICRA and etc.

Prof. Xuan Song’s main research interest are AI and its related research areas, such as data mining, intelligent system, especially on intelligent surveillance and information system design, mobility and spatio-temporal data mining. By now, he have published more than 100 technical publications in journals, book chapter, and international conference proceedings, including more than 40 high-impact papers in top-tier publications for computer science and robotics, such as ACM TOIS, ACM TIST, IEEE TPAMI, Applied Energy, IEEE Intelligent System, KDD, UbiComp, IJCAI, AAAI, ICCV, CVPR, ECCV, ICRA and etc. His research was featured in many Japanese and international media, including United Nations, the Discovery Channel, and Fast Company Magazine. He received Honorable Mention Award in UbiComp 2015.

Abstract

As the proliferation of IOT and Industry 4.0, the current intelligent sensing schemes have evolved from the traditional battery-based sensing to novel batteryless sensing. Take RFID as an example, by sufficiently leveraging the properties in backscatter-based communication, RFID system can be used to not only identify the objects but also locate and perceive the activities of the specified objects. In this talk, I will introduce our recent research progresses in the multi-modal battery-based sensing and RFID-based batteryless sensing. Specifically, we will show our research work in accurate localization, human activity sensing and human computer interaction with some video demos.

BIO

Prof. Lei Xie is a Professor and Ph.D. supervisor in the Department of Computer Science and Technology at Nanjing University. He is Young Chang Jiang Scholar. His research interests include RFID systems, pervasive and mobile computing, and Internet of things. He has published over 90 papers in ACM/IEEE TON, IEEE TMC, TPDS, IEEE TC, IEEE COMST, ACM TOSN, ACM MOBICOM, ACM UBICOMP, ACM MobiHoc, IEEE INFOCOM, IEEE ICNP, IEEE ICDCS, etc. He has published a monograph "RFID: Principles, Protocols and System Design" in Science Press. He received the first prize of Jiangsu Science and Technology Prize in 2016 and 2019, respectively. His research projects have been reported by a number of many high-end media, including CCTV, Jiangsu City Channel, SoHu, Sina, etc.

Abstract

Nowadays, traditional machine learning algorithms are suffered from the extreme scale of the modern data. Designing computational efficient machine learning algorithms for modern large-scale data set is very important. In this talk, I will introduce my work of using randomized hashing algorithms to scale up machine learning algorithms, including query recover for the commercial search engine, high dimensional anomaly detection, and Fast Bayesian Inferences. In the first part of the talk, I will first briefly introduce some background knowledge of Locality Sensitive Hashing. Then, I will introduce how to use Locality Sensitive Hashing (especially Minhash) to do fast string matching and query recovering for commercial search engines. In this work, we are able to do string matching in 44 million datasets within 2 milliseconds. Third, I will show how to use LSH structure to do very fast and memory efficient anomaly detection on high dimensional data. In this work, we are able to do very fast anomaly detection with less than 4 MB memory usage for more than 10GB of data. In the last part of my talk, I will introduce a novel use of LSH for speeding up Bayesian inference. This work is the first work that regards the LSH structure as a probabilistic sampler.

BIO

Dr. Chen Luo is an Applied Scientist at Amazon Search. He received his Ph.D. degree at Rice, working with Anshumali Shrivastava. He is working on Large-scale machine learning, including randomized hashing, System Mining, social network analysis. Before attending Rice, he was a master student in Key Laboratory of Symbolic Computation and Knowledge Engineering, Jilin University. He received his B.S degree from the Department of Computer Science, Jilin University. He has been working in MSRA in 2014 and in NEC Labs America in 2017 summer as a Research Intern. He published more than ten papers as the first author in top data mining and machine learning conferences and journals including KDD, WWW, AAAI, JMLR, etc.