A new algorithm for extracting a small representative subgraph from a very large graph

Many real-world networks are prohibitively large for data retrieval, storage and analysis of all of its nodes and links. Understanding the structure and dynamics of these networks entails creating a smaller representative sample of the full graph while preserving its relevant topological properties. In this report, we show that graph sampling algorithms currently proposed in the literature are not able to preserve network properties even with sample sizes containing as many as 20% of the nodes from the original graph. We present a new sampling algorithm, called Tiny Sample Extractor, with a new goal of a sample size smaller than 5% of the original graph while preserving two key properties of a network, the degree distribution and its clustering co-efficient. Our approach is based on a new empirical method of estimating measurement biases in crawling algorithms and compensating for them accordingly. We present a detailed comparison of best known graph sampling algorithms, focusing in particular on how the properties of the sample subgraphs converge to those of the original graph as they grow. These results show that our sampling algorithm extracts a smaller subgraph than other algorithms while also achieving a closer convergence to the degree distribution, measured by the degree exponent, of the original graph. The subgraph generated by the Tiny Sample Extractor, however, is not necessarily representative of the full graph with regard to other properties such as assortativity. This indicates that the problem of extracting a truly representative small subgraph from a large graph remains unsolved.

💡 Research Summary

The paper tackles the practical problem of analyzing massive networks—social media graphs, web graphs, biological interaction maps—where storing and processing the full adjacency structure is often infeasible. Traditional graph‑sampling techniques such as random‑walk crawls, metaseed‑based crawlers, edge‑sampling, and stratified sampling are evaluated and found to introduce substantial bias: even when 20 % of the original vertices are retained, key structural metrics (degree distribution, clustering coefficient, average path length) deviate noticeably from those of the full graph.

To overcome these limitations, the authors introduce the Tiny Sample Extractor (TSE), a novel algorithm whose primary goal is to produce a representative subgraph containing less than 5 % of the original nodes while faithfully preserving two fundamental properties: the degree distribution (specifically its power‑law exponent) and the average clustering coefficient. The core idea of TSE is to model and correct the measurement bias that arises during a crawling process. At each step of the exploration, the algorithm records the observed degree of the visited node and the connectivity of its neighborhood, compares these observations with the expected values derived from the current sample, and applies a bias‑compensating weight to the selection probability of subsequent nodes. This empirical correction drives the sampled subgraph toward the target degree exponent and clustering level.

The authors validate TSE on both synthetic scale‑free graphs and real‑world social networks (e.g., Facebook, Twitter). For each dataset, they compare TSE against the best known sampling baselines, measuring degree exponent, average clustering, average shortest‑path length, and assortativity. Results show that TSE consistently achieves a much tighter convergence to the original degree exponent and clustering coefficient, even when the sample size is as low as 3–5 % of the vertices. In contrast, the baseline methods require substantially larger samples (often >20 %) to reach comparable accuracy, and they still exhibit notable deviations. However, TSE does not reliably preserve assortativity or higher‑order community structure, highlighting a limitation of focusing on only two metrics.

The discussion emphasizes two key insights. First, explicitly estimating and correcting crawling bias is an effective strategy for small‑sample fidelity of low‑order topological statistics. Second, preserving a broader set of network characteristics likely requires a multi‑objective optimization framework or hybrid sampling schemes that can adaptively balance competing biases. The authors propose future work on learning‑based bias estimators and on integrating multiple sampling strategies to achieve a more holistic representation.

In conclusion, the Tiny Sample Extractor advances the state of graph sampling by demonstrating that a subgraph an order of magnitude smaller than previously thought necessary can still capture the essential degree distribution and clustering of the original network. This has immediate practical implications for reducing storage, transmission, and computational costs in large‑scale network analysis. Nevertheless, the paper acknowledges that extracting a truly representative small subgraph—one that simultaneously reflects assortativity, community structure, centrality, and other nuanced properties—remains an open challenge, inviting further research into bias‑aware, multi‑metric sampling methodologies.