[SIST Seminar] Parallel Graph Algorithm Design Based on Efficient Parallel Data Structures

ON2024-03-12TAG: ShanghaiTech UniversityCATEGORY: Lecture

Topic: Parallel Graph Algorithm Design Based on Efficient Parallel Data Structures

Speaker: WANG Letong, Department of Computer Science, University of California, Riverside (UCR)

Date and time: 10:00, March 13

Venue: Room 1A-200, SIST

Host: FAN Rui


Abstract:

Living  in the world of big data, we, as computer scientists, are challenged by  how to process the high volume of data efficiently. In most cases, we  want fast response time (e.g., training large AI models like ChatGPT),  as well as better accuracy within a certain budget of time (e.g.,  finding safer routes for autonomous driving). Parallel computing is a  key solution to enable these possibilities. With the prevalence of  parallel processors (e.g., multi-core CPUs, GPUs, and other  accelerators), designing efficient algorithms and software is now of  huge practical relevance and significant theoretical interest. This talk  is about a collection of parallel algorithms on graphs that take  advantage of novel parallel data structures. One is a parallel strong  connectivity algorithm based on faster reachability, which uses parallel  hash bags to avoid additional edge traversing. The other one is an algorithm to compute Influence Maximization on graphs, which uses parallel tree structures to keep track of the vertex with the maximum marginal influence.


Biography:

Letong Wang is a Ph.D. candidate in the Department of Computer Science at the University of California, Riverside (UCR). She received her B.E. in Computer Science and Engineering from ShanghaiTech University. Her research lies in the intersection of multi-core parallel algorithms and data science, with a focus on improving theoretical bounds and providing high-performance software. Her work has been published in the top conferences: SPAA, PPoPP, VLDB, SIGMOD, ECAI, and SSRR. Her paper “Provably Fast and Space-Efficient Parallel Biconnectivity” has been selected as Best Paper in PPoPP 23’.