Cyclosa: Redundancy-Free Graph Pattern Mining via Set Dataflowi
summary
Cyclosa implements a set dataflow prototype that rewrites pattern-centric mining into directed set operators (generator, combiner, reducer), runs on a single machine, and caches operator results in a fixed-size buffer with a two-phase priority check favoring high-degree vertices and early-flow results. Evaluation shows repeated set intersections account for over 80% of work and simple formula rewriting leaves >60% redundancies; Cyclosa reuses both explicit and implicit duplicates and outperforms GraphPi and SumPA by up to 16.28× and 5.52× respectively.
AI-written summary · about
chapters (13)
- 0:00Cyclosa introduction
- 2:23Limitations of existing systems
- 5:15Redundant computations in sets
- 7:23Set dataflow for redundancy
- 8:25Cyclosa architecture
- 9:32Set dataflow analyzer design
- 11:41Set dataflow representation
- 12:24Set-centric execution engine
- 14:12Memory management strategy
- 15:16Evaluation setup
- 16:39Caching and scalability
- 17:20Overheads and efficiency
- 18:23Conclusion
related talks
15mArya: Arbitrary Graph Pattern Mining with Decomposition-based Sampling
19mGraphs as Streams: Rethinking Graph Processing in ... #bbuzz
25mUsing Set Cover to Optimize a Large-Scale Low Latency Distributed Graph
47mPattern Search (Data Mining: Pattern Discovery)
41mIntroducing Cascalog: Functional Data Processing for Hadoop
24mLarge Scale Graph Computation with Apache Giraph
56mGraph Analytic Systems
20mGraphWalker: An I/O-Efficient and Resource-Friendly Graph Analytic System...