"Harmonizing High-Level Abstraction and High Performance for Graph Mining"
Graph mining algorithms that aim at identifying structural patterns in graphs are typically more complex than graph computation algorithms such as breadth first search. Researchers have implemented several systems with high-level and flexible interfaces customized for tackling graph mining problems. However, we find that for triangle counting, one of the simplest graph mining problems, such systems can be several times slower than a single-threaded implementation of a straightforward algorithm. In this talk, I will reveal the root causes of the severe inefficiency of state-of-the-art graph mining systems and the challenges to address the performance problems. I will describe AutoMine, a system we developed to compile arbitrary patterns to efficient C++ code.
Bo Wu is an Associate Professor in the Department of Computer Science at Colorado School of Mines. His research focuses on leveraging compiler and runtime techniques to build efficient software systems for large-scale graph analytics and machine learning applications on heterogeneous platforms. He received the best paper award at SC’15, an NSF CRII Award, an NSF Early Career Award, and an NSF SPX Award.