I am a postdoctoral researcher in Caltech Computing and Mathematical Sciences. Before that, I received my Ph.D. degree in Computer Science from University of Southern California.
I work on machine learning for time series/spatiotemporal data and its applications, especially in the field of computational sustainability. I am generally interested in sequential decision making, optimization and spatiotemporal modeling. My work has been applied to intelligent transportation, climate informatics, and social media anomaly detection. For more details, see my curriculum vitae and research statement.
In 2018 fall, I will be joining Northeastern College of Computer and Information Science as an assistant professor. And I am looking for talented students who are excited about machine learning!
Contact Info
- E-mail: qiyu [at] caltech [dot] edu.
- Address: 330 S Chester Ave, Pasadena, CA 91125
News
- July 2017: Give talks at Peking University, DiDi Chuxing, JiaoTong University and Zhejiang University !
- June 2017: Co-organize time series workshop at ICML 2017, please submit!
- May 2017: New paper on Graph Convolutional Recurrent Neural Network for Spatiotemporal Forecasting!
- April 2017: SDM Best paper runner-up for Deep Learning for Traffic Forecasting!
- April 2017: Won ACM scholarship for Turing Award Celebration!
Projects
Tensor Learning for Spatiotemporal Analysis
A scalable framework for analyzing spatiotemporal data
Large-scale spatiotemporal data such as climate measurements, location-based social networks have complex structures. In this project, we develop a scalable framework based on low-rank tensor learning to analyze them. We propose a series of algorithms ranging from offline, online to memory efficient learning to solve the problem. We show that our framework can produce accurate predictions in a scalable way and discover interesting patterns in spatiotemporal data.
ICML 2016 ICML 2015 NIPS 2014
Spatiotemporal Models for Traffic Prediction
A spatiotemporal approach to predict road network traffic
Given a series of road network snapshots, we propose a latent space model to capture its topological and temporal properties. We present an incremental online algorithm which sequentially and adaptively learns the spatiotemporal patterns. Our framework enables real-time traffic prediction on a large volumne of Los Angeles traffic sensor data.
SDM 2017 KDD 2016
Social Media Group Anomaly Detection
A probabilistic model to detect anomalous groups in social media
Traditional anomaly detection techniques focus on individual point anomalies. With the recent popularity of social media, new types of anomalous behaviors arise. We study the collective behavior of individuals and detect group anomalies. We take a generative approach by proposing a hierarchical Bayes model. Our approach provides a robust and interpretable way of discovering latent groups and detecting group anomalies.
KDD Exploration 2016 TKDD 2015 KDD 2014
Publications
Preprints
Conferences