报告题目: Multimodal Learning on Spatiotemporal Data
报 告 人：张超 博士
主 持 人：张伟 副研究员
报告时间：2018年6月20日 周三 15:30-16:30
报告地点：中北校区数学馆201报告厅

报告人简介：

Chao Zhang is a PhD candidate in Computer Science at University of Illinois at Urbana-Champaign, advised by Professor Jiawei Han. He will be joining School of Computational Science and Engineering at Georgia Tech as an assistant professor in December 2018. His research lies in data mining and machine learning, with a focus on mining knowledge from text and spatiotemporal data. Chao has published more than 30 papers in top-tier conferences and journals, including KDD, WWW, SIGIR, VLDB, AAAI and TKDE. He is the recipient of the ECML/PKDD Best Student Paper Runner-up Award (2015) and the Chiang Chen Overseas Graduate Fellowship (2013). His developed technologies have received multiple media coverages and been transferred to research institutes and industrial companies. More information can be found at http://chaozhang.org/

 
报告摘要：
Text and spatiotemporal data are converging in many domains, such as urban science, business, scientific research, and healthcare. Effective and scalable analytics of text-rich spatiotemporal data can be game changing for various real-life applications. However, such data pose great challenges to existing data mining research since they are unstructured, fragmented, and multimodal in nature. In this talk, I will introduce a multimodal learning framework that harnesses the power of text-rich spatiotemporal data. The framework consists of three modules for addressing several key challenges in mining text-rich spatiotemporal data, including: (1) multimodal structuring---how to organize unstructured text-rich spatiotemporal data to facilitate on-demand knowledge discovery? (3) multimodal pattern discovery---how to discover interesting patterns in the multimodal space? and (3) multimodal prediction---how to integrate different modalities to make accurate predictions?  With these three modules, the framework is capable of turning text-rich spatiotemporal data into a structured and multidimensional knowledge cube, which can serve as a versatile and easy-to-use knowledge engine for many applications. Finally, I will share some future research directions on leveraging text-rich spatiotemporal data for building next-generation intelligent systems.