[cs-talks] CS Upcoming Seminars: PhD Thesis Defense (Mon) + BUSec (Wed) + IVC Sem (Thurs)
fgreen1 at bu.edu
Mon Feb 1 10:21:30 EST 2016
PhD Thesis Defense
Sequence Queries on Temporal Graphs
Haohan Zhu, BU
Monday, Febraury 1, 2016 at 12pm in MCS 148
Abstract: Graphs that evolve over time are called temporal graphs. They can be used to describe and represent real-world networks, including transportation networks, social networks, and communication networks, with higher fidelity and accuracy. However, research is still limited on how to manage large scale temporal graphs and execute queries over these graphs efficiently and effectively. This thesis investigates the problems of temporal graph data management related to node and edge sequence queries. In temporal graphs, nodes and edges can evolve over time. Therefore, sequence queries on nodes and edges can be key components in managing temporal graphs. In this thesis, the node sequence query decomposes into two parts: graph node similarity and subsequence matching. For node similarity, this thesis proposes a modified tree edit distance that is metric and polynomially computable and has a natural, intuitive interpretation. Note that the proposed node similarity works even for inter-graph nodes and therefore can be used for graph de-anonymization, network transfer learning, and cross-network mining, among other tasks. The subsequence matching query proposed in this thesis is a framework that can be adopted to index generic sequence and time-series data, including trajectory data and even DNA sequences for subsequence retrieval. For edge sequence queries, this thesis proposes an efficient storage and optimized indexing technique that allows for efficient retrieval of temporal subgraphs that satisfy certain temporal predicates. For this problem, this thesis develops a lightweight data management engine prototype that can support time-sensitive temporal graph analytics efficiently even on a single PC.
George Kollios (Advisor)- Computer Science, Boston University
Evimaria Terzi- Computer Science, Boston University
Mark Crovella- Computer Science, Boston University
Stan Sclaroff- Computer Science, Boston University
Hongwei Xi (Chair)- Computer Science, Boston University
Cryptography in the Application Layer: Establishing an Affordable DDoS Defense over Untrusted Clouds
Yossi Gilad, BU
Wednesday, Febraury 3, 2016 at 9:45am in MCS 180
Abstract: We present clientless secure objects, a new mechanism to secure web-content against adversarial cache and proxy systems. Deploying clientless secure objects avoids trusting these systems with private keys or user-data, yet does not require changing existing clients or cache/proxy services. We then investigate applications of clientless secure objects and present CDN-on-Demand, a software-based defense that deploys proxy servers on less expensive and less trusted clouds to minimize costs. Administrators of small to medium websites can install to resist powerful DDoS attacks with a fraction of the cost of comparable commercial CDN services. Upon excessive load, CDN-on-Demand serves clients from a scalable set of proxies that it automatically deploys on multiple IaaS cloud providers. Joint work with Michael Goberman, Amir Herzberg and Michael Sudkovich
Collective Insight: Crowd-driven Image Understanding
Genevieve Patterson, Brown University
Thursday, February 4, 2016 at 2pm in MCS 180
Abstract: Crowd annotated datasets have become a mainstay in computer vision, enabling some of the most significant discoveries of recent years. However, the research community has yet to fully exploit the full range of human intelligence available in the crowd. This talk demonstrates that, by going beyond naive annotation, researchers can access the vast potential of crowdsourcing. Using polling, active learning, intelligent annotation protocols, and other techniques we are able to leverage the crowd to discover a taxonomy of visual attributes, build detectors with minimal supervision, and economically label massive datasets.
Bio: Genevieve is a PhD Candidate in Computer Vision at Brown University. Her work on crowd-driven visual classification was recently awarded runner-up for Best Paper at the AAAI Conference on Human Computation (HCOMP). She built and maintains the SUN Attribute dataset, a widely used resource for scene understanding. She continues to work on attribute and scene understanding.
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the cs-talks