A GPU-based Graph Pattern Mining System.

Graph pattern mining (GPM) is getting increasingly important in recent years. Many graph pattern mining frameworks try to use universal primitives to deal with various graph pattern mining tasks. However, most of them suffer from unsatisfactory performance because of the exponential complexity of GPM. GPU is a new hardware with great parallelism, and many graph algorithms have achieved significant performance improvements on GPU. In this demo, we propose a graph pattern mining framework on GPU, called GAMMA. GAMMA proposes effective and flexible interfaces for users to implement their mining tasks conveniently. GPM has extensive intermediate results in parallel environments. We make full use of host memory to deal with large-scale graphs and extensive intermediate results. We also present several optimizations to process large graphs. GAMMA has great scalability and performance advantages compared with state-of-the-art graph mining works in experiments.