Prof. Eric Xing from CMU

We are pleased that Professor Eric Xing from CMU visited McGill and our lab.

Dr. Eric Xing is a Professor of Machine Learning in the School of Computer Science at Carnegie Mellon University, and Director of the CMU/UPMC Center for Machine Learning and Health. His principal research interests lie in the development of machine learning and statistical methodology, and large-scale computational system and architecture; especially for solving problems involving automated learning, reasoning, and decision-making in high-dimensional, multimodal, and dynamic possible worlds in artificial, biological, and social systems. Professor Xing received a Ph.D. in Molecular Biology from Rutgers University, and another Ph.D. in Computer Science from UC Berkeley. He servers (or served) as an associate editor of the Annals of Applied Statistics (AOAS), the Journal of American Statistical Association (JASA), the IEEE Transaction of Pattern Analysis and Machine Intelligence (PAMI), the PLoS Journal of Computational Biology, and an Action Editor of the Machine Learning Journal (MLJ), the Journal of Machine Learning Research (JMLR). He was a member of the DARPA Information Science and Technology (ISAT) Advisory Group, a recipient of the NSF Career Award, the Sloan Fellowship, the United States Air Force Young Investigator Award, and the IBM Open Collaborative Research Award. He served as the Program Chair of ICML 2014.

Dr. Pieter Johannes Mosterman visit

We are pleased that Dr. Pieter Johannes Mosterman from MathWorks visited our lab.

Pieter Johannes Mosterman is a Senior Research Scientist at MathWorks in Natick, Massachusetts. He also holds an Adjunct Professorship at the School of Computer Science at McGill University in Montreal, Canada. His primary research interests are in Computer Automated Multi-paradigm Modeling with principal applications in design automation, training systems, and fault detection, isolation, and reconfiguration.



Research and Projects

Sustainable Datacenters

The computing capacity and scale of data centers are increasing to meet the soaring demand for IT applications and services. Mega data center can host thousands of servers and require up to tens of megawatts of electricity. The high power consumption causes two serious consequences. First, data center operators may face millions of dollars annually charged by the electrical grid. Second, the enormous energy consumption can lead to negative environmental impacts. This project focuses on data center sustainability, to cut operational cost and increase renewable integration for data centers.
Selected publications:
  • F. Kong and X. Liu, "GreenPlanning: Optimal Energy Source Selection and Capacity Planning for Green Datacenters", in the ACM/IEEE 7th International Conference on Cyber-Physical Systems (ICCPS), 2016, pp. 1-10. (Acceptance rate: 27.8%)
  • F. Kong and X. Liu, "A Survey on Green-Energy-Aware Power Management for Datacenters", in ACM Computing Surveys (CSUR), 2015, pp. 30:1-30:38. (IF: 3.37).
  • A. Rahman, X. Liu and F. Kong, "A Survey on Geographic Load Balancing based Data Center Power Management in the Smart Grid Environment", in IEEE Communications Surveys and Tutorials (COMST), 2014, pp. 214-233. (IF: 6.81).
  • X. Liu and F. Kong, "Datacenter Power Management in Smart Grids", in Foundations and Trends in Electronic Design Automation, now publishers Inc., 2015, pp. 1-98. (Monograph)
  • F. Kong, X. Lu, M. Xia, X. Liu and H. Guan, "Distributed Optimal Datacenter Bandwidth Allocation for Dynamic Adaptive Video Streaming", in the 23rd ACM Multimedia (MM), 2015, pp. 531-540. (Full paper, acceptance rate: 20.6%)
  • F. Kong, X. Liu and L. Rao, "Optimal Energy Source Selection and Capacity Planning for Green Datacenters", in the 40th ACM International Conference on Measurement and Modeling of Computer Systems (SIGMETRICS), 2014, pp. 575-576. (Extended abstract)
  • Z. Sun, F. Kong, X. Liu, X. Zhou and X. Chen, "Intelligent Joint Spatio-temporal Management of Electric Vehicle Charging and Data Center Power Consumption", in the 5th International Green Computing Conference (IGCC), 2014, pp. 1-8.
  • C. Dong, F. Kong, X. Liu and H. Zeng, "Green Power Analysis for Geographical Load Balancing Based Datacenters", in the 4th International Green Computing Conference (IGCC), 2013, pp. 1-8.

EV: Green Charging

The power flow incurred by electric vehicle(EV) and renewable energy are both crucial to the future smart grid. Yet how to integrated them into power systems remains largely unexplored. In the Cyber-Physical Intelligence Laboratory at McGill University, we are working on developing a series of innovative technology and market strategies to achieve this integration in an efficient, reliable and real-time manner.
Selected publications:
  • Q. Wang, X. Liu, J. Du, and F. Kong, "Smart Charging for Electric Vehicles: A Survey From the Algorithmic Perspective," IEEE Communications Surveys Tutorials, vol. 18, no. 2, pp. 1500–1517, Secondquarter 2016.
  • F. Kong, Q. Xiang, L. Kong, and X. Liu, "On-Line Event-Driven Scheduling for Electric Vehicle Charging via Park-and-Charge", accepted by the 36th IEEE Real-Time Systems Symposium (RTSS), 2016. (Acceptance rate: 23.4%)
  • F. Kong, X. Liu, Z. Sun, and Q. Wang, "Smart Rate Control and Demand Balancing for Electric Vehicle Charging", in the ACM/IEEE 7th International Conference on Cyber-Physical Systems (ICCPS), 2016, pp. 1-10. (Acceptance rate: 27.8%)
  • F. Kong and X. Liu, "Distributed Deadline and Renewable Aware Electric Vehicle Demand Response in the Smart Grid", in the 36th IEEE Real-Time Systems Symposium (RTSS), 2015, pp. 23-32. (Acceptance rate: 22.5%)
  • F. Kong, C. Dong, X. Liu and H. Zeng, "Quantity vs Quality: Optimal Harvesting Wind Power for the Smart Grid", in Proceedings of the IEEE (PIEEE), 2014, pp. 1762-1776. (IF: 4.93).
  • F. Kong, C. Dong, X. Liu and H. Zeng, "Blowing Hard Is Not All We Want: Quantity vs Quality of Wind Power in the Smart Grid", in 33rd Annual IEEE International Conference on Computer Communications (INFOCOM), 2014, pp. 2813-2821. (Acceptance rate: 19.5%)
  • Z. Sun, F. Kong, X. Liu, X. Zhou and X. Chen, "Intelligent Joint Spatio-temporal Management of Electric Vehicle Charging and Data Center Power Consumption", in the 5th International Green Computing Conference (IGCC), 2014, pp. 1-8.

VSmart: DSRC-based Smart Vehicle Testbed

VSmart is a DSRC-enabled smart vehicle testbed established at Cyber-Physical Intelligence Laboratory (CPS-Lab), McGill University, and is supported by NSERC and General Motors Company. VSmart explores and illustrates the potential to enhance driving safety and traffic efficiency with V2V communications (especially DSRC).

Selected Publications

  • L. Kong, X. Chen, X. Liu, L. Rao. "FINE: Frequency-divided Instantaneous Neighbors Estimation System in Vehicular Networks". IEEE PerCom Concise Paper, St. Louis, Missouri, USA, 2015.

AppAudit: Detecting Data-leaking Mobile Apps

We design AppAudit, a program analysis framework that checks if an Android application leaks sensitive personal data. AppAudit is designed with minimalism, using least possible memory and least amount of time. Current prototype could vet a real app with 256MB memory in 5 seconds on average. AppAudit can be used for three use cases:
  • mobile app developers could use AppAudit to check if their apps include any data-leaking libraries or modules
  • the app market could use AppAudit to vet newly uploaded apps and remove data-leaking ones
  • mobile users could use AppAudit to avoid installing data-leaking apps

Selected publications:
  • M. Xia, L. Gong, Y. Lv, Zh. Qi, X. Liu: Effective Real-time Android Application Auditing. The 36th IEEE Symposium on Security and Privacy (IEEE S&P'15)

Apps Drain Battery Because of Memory Leaks

Mobile operating systems embrace new mechanisms that reduce energy consumption for common usage scenarios. The background app design is a repre sentative implemented in all major mobile OSes. The OS keeps apps that are not currently inter acting with the user in memory to avoid repeated app loading. This mechanism improves responsive ness and reduces the energy consumption when the user switches apps. However, we demonstrate that application errors, in particular memory leaks that cause system memory pressure, can easily cripple this mechanism. In this paper, we conduct experi ments on real Android smartphones to 1) evaluate how the background app design improves respon siveness and saves energy; 2) characterize memory leaks in Android apps and outline its energy im pact; 3) propose design improvements to retrofit the mechanism against memory leaks.
Selected publications:
  • M. Xia, W. He, X. Liu, J. Liu: Why Application Errors Drain Battery Easily? A Study of Memory Leaks in Smartphone Apps. The 5th Workshop on Power-Aware Computing and Systems (HotPower'13)

Selected Publications

  • L. Kong, L. He, X. Liu, Y. Gu, M. Wu, X. Liu. "Privacy-Preserving Compressive Sensing for Crowdsensing based Trajectory Recovery". IEEE ICDCS, Columbus, Ohio, USA, 2015.

mZig: Multi-Packet Reception in ZigBee

mZig is a novel physical layer design that enables a receiver to simultaneously decode multiple packets from different transmitters in ZigBee. As a low-power and low-cost wireless protocol, the promising ZigBee has been widely used in sensor networks, cyber-physical system, and smart buildings. Since ZigBee based networks usually adopt tree or cluster topology, the convergecast scenarios are common in which multiple transmitters need to send packets to one receiver. For example, in a smart home, all appliances report data to one control plane via ZigBee. However, concurrent transmissions lead to the severe collision problem. The conventional ZigBee avoids collisions using backoff time, which introduces additional time overhead. Advanced methods resolve collisions instead of avoidance, in which the state-of-the-art ZigZag resolves one m-packet collision requiring m retransmissions.
Selected publications:
  • L. Kong, X. Liu. "mZig: Enabling Multi-Packet Reception in ZigBee". ACM MobiCom, Paris, France, 2015.

Performance Optimization and Tuning of 802.11 Wifi

Selected publication:
  • A. J. Pyles, X. Qi, G. Zhou, M. Keally, and X. Liu, "SAPSM: Smart Adaptive 802.11 PSM for Smartphones," in Proceedings of the 2012 ACM Conference on Ubiquitous Computing, New York, NY, USA, 2012, pp. 11–20. (Acceptance rate = 19.3%, 58 out of 301).
  • X. Xing, J. Dang, S. Mishra, and X. Liu, "A highly scalable bandwidth estimation of commercial hotspot access points," in 2011 Proceedings IEEE INFOCOM, 2011, pp. 1143–1151.
  • A. J. Pyles, Z. Ren, G. Zhou, and X. Liu, "SiFi: Exploiting VoIP Silence for WiFi Energy Savings Insmart Phones," in Proceedings of the 13th International Conference on Ubiquitous Computing, New York, NY, USA, 2011, pp. 325–334. (Acceptance rate = 16.6%, 50 out of 302).
  • X. Xing, S. Mishra, and X. Liu, "ARBOR: Hang Together Rather Than Hang Separately in 802.11 WiFi Networks," in 2010 Proceedings IEEE INFOCOM, 2010, pp. 1–9. (Acceptance rate = 17.5%, 276 out of 1575).

Aerial - Device-free activity recognition

Aerial is a technology that senses and distinguishes who you are, where you are and what you are doing using Wi-Fi signals already present in your house. With just a single device, you can monitor your entire home. No need to buy a variety of expensive sensors and invasive cameras. Right out of the box, the aerial cube is packed with features to help you become more aware of what is happening in your home. As long as there are standard Wi-Fi signals in the air, aerial is good to go.

Cloud Computing

Selected publications:
  • Y. Hua, B. Xiao, and X. Liu, "NEST: Locality-aware approximate query service for cloud computing," in 2013 Proceedings IEEE INFOCOM, 2013, pp. 1303–1311. (Acceptance rate=17%, 280 out of 1613).
  • Y. Hua, X. Liu, and D. Feng, "Neptune: Efficient remote communication services for cloud backups," in 2014 Proceedings IEEE INFOCOM, 2014, pp. 844–852. (Acceptance rate=19.4%, 320 out of 1645).
  • Y. Hua, X. Liu, and H. Jiang, "AN℡OPE: A Semantic-Aware Data Cube Scheme for Cloud Data Center Networks," IEEE Transactions on Computers, vol. 63, no. 9, pp. 2146–2159, Sep. 2014.
  • Y. Hua, X. Liu, and D. Feng, "Data Similarity-Aware Computation Infrastructure for the Cloud," IEEE Transactions on Computers, vol. 63, no. 1, pp. 3–16, 2014.

Big Data Processing and Machine Learning Systems and Applications

Update-Efficient and Parallel-Friendly Content-based Indexing System
The sheer volume of contents generated by today's Internet services are stored in the cloud. The effective indexing method is important to provide the content to users on demand. The indexing method associating the user-generated metadata with the content is vulnerable to the inaccuracy caused by the low quality of the metadata. While the content-based indexing does not depend on the error-prone metadata, the state-of-the-art research focuses on developing descriptive features and miss the system-oriented considerations when incorporating these features into the practical cloud computing systems.
We propose an Update-Efficient and Parallel-Friendly content-based indexing system, called Partitioned Hash Forest (PHF). The PHF system incorporates the state-of-the-art content-based indexing models and multiple system-oriented optimizations. PHF contains an approximate content-based index and leverages the hierarchical memory system to support the high volume of updates. Additionally, the content-aware data partitioning and lock-free concurrency management module enable the parallel processing of the concurrent user requests. We evaluate PHF in terms of indexing accuracy and system efficiency by comparing it with the state-of-the-art content-based indexing algorithm and its variances. We achieve the significantly better accuracy with less resource consumption, around 37% faster in update processing and up to 2.5X throughput speedup in a multi-core platform comparing to other parallel-friendly designs.
Distributed (Deep) Machine Learning Community (DMLC)
DMLC is a community of awesome distributed machine learning projects, including the well-known parallel gradient boost tree model XGBoost, and the deep learning system, MXNet, etc.
XGBoost is an optimized distributed gradient boosting library designed to be highly efficient, flexible and portable. It implements machine learning algorithms under the Gradient Boosting framework. It has been the winner solution of many Kaggle machine learning competitions.
Nan Zhu, a member of CPSLAB, is leading the efforts on the development of XGBoost jvm-packages and serves as the committee member of DMLC. The main goal of XGBoost jvm-packages is to achieve seamless integration between XGBoost and JVM-based parallel data processing systems like Apache Spark. With integration, users can enjoy both the convenient interfaces in systems like Spark and the high performance of XGBoost.
You can check the release blog.
Selected Publications:
  • N. Zhu, L. Rao, X. Liu, "PD2F: Running a Parameter Server within a Distributed Dataflow Framework," Workshop on Machine Learning Systems at Neural Information Processing Systems (NIPS), 2015.
  • M. Xia, N. Zhu, S. Elnikety, X. Liu, and Y. He, "Performance Inconsistency in Large Scale Data Processing Clusters," presented at the Proceedings of the 10th International Conference on Autonomic Computing (ICAC 13), 2013, pp. 297–302.

Realtime Systems and Applications

  • W. Dong, C. Chen, J. Bu, X. Liu, and Y. Liu, "D2: Anomaly Detection and Diagnosis in Networked Embedded Systems by Program Profiling and Symptom Mining," in 34th IEEE International Real-Time Systems Symposium (RTSS), 2013, pp. 202–211.
  • W. Liu, M. Yuan, X. He, Z. Gu, and X. Liu, "Efficient SAT-Based Mapping and Scheduling of Homogeneous Synchronous Dataflow Graphs for Throughput Optimization," in 29th IEEE International Real-Time Systems Symposium, 2008, pp. 492–504.
  • W. Liu, M. Yuan, X. He, Z. Gu, and X. Liu, "Efficient SAT-Based Mapping and Scheduling of Homogeneous Synchronous Dataflow Graphs for Throughput Optimization," in 29th IEEE International Real-Time Systems Symposium (RTSS), 2008, pp. 492–504.
  • J. Heo, D. Henriksson, X. Liu, and T. Abdelzaher, "Integrating Adaptive Components: An Emerging Challenge in Performance-Adaptive Systems and a Server Farm Case-Study," in 28th IEEE International Real-Time Systems Symposium (RTSS), 2007, pp. 227–238.
  • Q. Wang, X. Liu, J. Hou, and L. Sha, "GD-Aggregate: A WAN Virtual Topology Building Tool for Hard Real-Time and Embedded Applications," in 28th IEEE International Real-Time Systems Symposium (RTSS), 2007, pp. 379–388.
  • X. Liu and T. Abdelzaher, "On Non-Utilization Bounds for Arbitrary Fixed Priority Policies," in 12th IEEE Real-Time and Embedded Technology and Applications Symposium (RTAS'06), 2006, pp. 167–178. (Best Paper Award Finalist).
  • L. Sha, X. Liu, Y. Lu, and T. Abdelzaher, "Queueing model based network server performance control," in 23rd IEEE International Real-Time Systems Symposium (RTSS), 2002, pp. 81–90.

Sensor Networks

  • L. Kong, Q. Xiang, X. Liu, X. Liu, X. Gao, G. Chen, M. Wu. "ICP: Instantaneous Clustering Protocol for Wireless Sensor Networks". Elsevier Computer Networks (COMNET),Vol. 101, pp. 144-157, 2016.
  • L. Kong, M. Xia, X. Y. Liu, M. Y. Wu, and X. Liu, "Data loss and reconstruction in sensor networks," in 2013 Proceedings IEEE INFOCOM, 2013, pp. 1654–1662. (Acceptance rate=17%, 280 out of 1613).
  • Y. Gao, W. Dong, C. Chen, J. Bu, G. Guan, X. Zhang, and X. Liu, "Pathfinder: Robust path reconstruction in large scale sensor networks with lossy links," in 2013 21st IEEE International Conference on Network Protocols (ICNP), 2013, pp. 1–10. (Acceptance rate=18%, 46 out of 251).
  • W. Dong, Y. Liu, C. Wang, X. Liu, C. Chen, and J. Bu, "Link quality aware code dissemination in wireless sensor networks," in 2011 19th IEEE International Conference on Network Protocols, 2011, pp. 89–98. (Acceptance rate = 16.4%, 31 out of 189).
  • X. Liu and T. Abdelzaher, "Nonutilization Bounds and Feasible Regions for Arbitrary Fixed-priority Policies," ACM Trans. Embed. Comput. Syst., vol. 10, no. 3, p. 31:1–31:25, May 2011.
  • W. He, X. Liu, H. V. Nguyen, K. Nahrstedt, and T. Abdelzaher, "PDA: Privacy-Preserving Data Aggregation for Information Collection," ACM Trans. Sen. Netw., vol. 8, no. 1, p. 6:1–6:22, Aug. 2011.

Driving Safety Application

A large number of car accidents occur at intersections every year mainly due to drivers’ "illegal maneuver" or "unsafe behavior". To promote traffic safety, we present SafeCam, a smartphone-based system that jointly leverages vehicle dynamics and the real-time traffic control information (e.g., traffic signals) to detect and study driver dangerous behaviors at intersections. In particular, SafeCam uses embedded sensors (i.e., inertial sensors) on the phone to generate soft hints tracking different driving conditions while at the same time adopts vision-based algorithms to recognize intersection-related critical driving events including unsafe turns, running stop signs and running red lights. In order to improve the system efficiency, we utilize adaptive color filtering under two lighting conditions (e.g., sunny and cloudy) and deploy the subsampling methods to make a trade off between the detection rate and the processing latency.
Selected publications:
  • L. Jiang, X. Chen, and W. He, "SafeCam: Analyzing intersection-related driver behaviors using multi-sensor smartphones," in 2016 IEEE International Conference on Pervasive Computing and Communications (PerCom), 2016, pp. 1–9.

A Study of Facebook Likes

This is a project revealing the flaws of the Facebook Like system. These flaws widely exist in most of the websites that take use of the Facebook Like button, and can be exploited by spammers to automatically generate large amount of fake Likes for profits. Meanwhile, these flaws make legitimate users unintentionally generate Facebook Likes against the online contents they don't like or even have negative feelings about. These flaws endanger the Facebook Like ecosystem and the benefits of both legitimate users and advertisers. Many famous websites, such as FoxNews, abcNews, ESPN, HuffingtonPost etc., are victims of these flaws.

SafeVChat: Safety and Security in Online Video Chat Systems

Online video chat systems such as Chatroulette have become increasingly popular as a way to meet and converse one-on-one via video and audio with other users online in an open and interactive manner. At the same time, safety and security concerns inherent in such communication have been little explored. Our research group at University of Colorado, Boulder, USA and McGill University, Montreal, Canada seeks to conduct ground-breaking research in the context of an online video chat system.
Selected publications:
  • Y.-L. Liang, X. Xing, H. Cheng, J. Dang, S. Huang, R. Han, X. Liu, Q. Lv, and S. Mishra, “SafeVchat: A System for Obscene Content Detection in Online Video Chat Services,” ACM Transactions on Internet Technology, vol. 12, no. 4, p. 13:1–13:26, Jul. 2013.
  • X. Xing, Y.-L. Liang, H. Cheng, J. Dang, S. Huang, R. Han, X. Liu, Q. Lv, and S. Mishra, “SafeVchat: Detecting Obscene Content and Misbehaving Users in Online Video Chat Services,” in Proceedings of the 20th International Conference on World Wide Web (WWW'11), New York, NY, USA, 2011, pp. 685–694.

ISC: Adult Account Detection on Twitter

The widespread of adult content on online social networks (e.g., Twitter) is becoming an emerging yet critical problem. An automatic method to identify accounts spreading sexually explicit content (i.e., adult account) is of significant values in protecting children and improving user experiences. Traditional adult content detection techniques are ill-suited for detecting adult accounts on Twitter due to the diversity and dynamics in Twitter content. In this article, we formulate the adult account detection as a graph based classification problem and demonstrate our detection method on Twitter by using social links between Twitter accounts and entities in tweets. As adult Twitter accounts are mostly connected with normal accounts and post many normal entities, which makes the graph full of noisy links, existing graph based classification techniques cannot work well on such a graph. To address this problem, we propose an iterative social based classifier (ISC), a novel graph based classification technique resistant to the noisy links. Evaluations using large-scale real-world Twitter data show that, by labeling a small number of popular Twitter accounts, ISC can achieve satisfactory performance in adult account detection, significantly outperforming existing techniques.
Selected publications:
  • H. Cheng, X. Xing, X. Liu, Q. Lv, "ISC: An Iterative Social Based Classifier for Adult Account Detection on Twitter", IEEE Transactions on Knowledge & Data Engineering, vol.27, no. 4, pp. 1045-1056, April 2015.