Publications

Arya introduces a decomposition-based sampling framework that supports interactive mining of arbitrary subgraph patterns on production-scale graphs. The system decomposes complex pattern queries into reusable sampling plans that deliver unbiased estimates with theoretical accuracy guarantees and low latency. Experiments on industry workloads show that Arya answers complex pattern queries orders of magnitude faster than prior motif enumeration systems while retaining accuracy.

Extends the prior study to focus on the number 3, compares its properties with earlier findings, and motivates investigating the number 4 as the next step.

Reviews properties of the number 2, evaluates its behavior in different contexts, and identifies extending the framework to the number 3 as the next challenge.