Fast and Accurate Random Walk with Restart on Dynamic Graphs with Guarantees

Abstract

Given a time-evolving graph, how can we track similarity between nodes in a fast and accurate way, with theoretical guarantees on the convergence and the error? Random Walk with Restart (RWR) is a popular measure to estimate the similarity between nodes and has been exploited in numerous applications. Many real-world graphs are dynamic with frequent insertion/deletion of edges; thus, tracking RWR scores on dynamic graphs in an efficient way has aroused much interest among data mining researchers. Recently, dynamic RWR models based on the propagation of scores across a given graph have been proposed, and have succeeded in outperforming previous other approaches to compute RWR dynamically. However, those models fail to guarantee exactness and convergence time for updating RWR in a generalized form.In this paper, we propose OSP, a fast and accurate algorithm for computing dynamic RWR with insertion/deletion of nodes/edges in a directed/undirected graph. When the graph is updated, OSP first calculates offset scores around the modified edges, propagates the offset scores across the updated graph, and then merges them with the current RWR scores to get updated RWR scores. We prove the exactness of OSP and introduce OSP-T, a version of OSP which regulates a trade-off between accuracy and computation time by using error tolerance ϵ. Given restart probability c, OSP-T guarantees to return RWR scores with O(ϵ/c) error in O(log (1−c) ( ϵ 2 )) iterations. Through extensive experiments, we show that OSP tracks RWR exactly up to 4605× faster than existing static RWR method on dynamic graphs, and OSP-T requires up to 15× less time with 730× lower L1 norm error and 3.3× lower rank error than other state-of-the-art dynamic RWR methods.

Optimizing Network Propagation for Multi-Omics Data Integration

“…Whereas the Laplacian is based on subtracting the degree matrix D, the degree row-normalized adjacency matrix is obtained by multiplying the inverse of D with A. We normalize the matrix row-wise (Yoon, Jin, and Kang 2018;Lovász 1993;Sonnenberg 2009) and define the degree row-normalized adjacency matrix as…”

Fast and Accurate Anomaly Detection in Dynamic Graphs with a Two-Pronged Approach

“…This equation is interpreted as a propagation of scores across a graph: initial scores in the starting vector b are propagated across the graph by multiplying withà ⊤ ; since the damping factor c is smaller than 1, propagated scores converge, resulting in PageRank scores. As shown in [28], PageRank computation time is proportional to the L1 length of the starting vector b, since small L1 length of b leads to faster convergence of iteration ( ∞ i=0 (cà ⊤ ) i ) with damping factor c. Here L1 length of a vector is defined as the sum of absolute values of its entries. The L1 length of a matrix is defined as the maximum L1 length of its columns.…”

“…From now on, denote ∆A = B ⊤ −Ã ⊤ , the difference between transpose of normalized matrices A ⊤ andB ⊤ . Lemma 1 (Dynamic PageRank, Theorem 3.2 in [28]). Given updates ∆A in a graph during ∆t, an updated PageRank vector p(t + ∆t) is computed incrementally from a previous PageRank vector p(t) as follows:…”

A Survey on Personalized PageRank Computation Algorithms

“…They update the precomputed PPR vectors when the edges are added or deleted, and they also can be categorized into major approaches like other works. IRWR [49] updates PPR vectors based on direct solving formula, and LayFwdUpdate [83], and OSP [84] are dynamic variants of iterative methods. TrackingPPR [85] works on the bookmark coloring scheme.…”