Fast-PPR

Lofgren, Peter; Banerjee, Siddhartha; Goel, Ashish; Seshadhri, C.

doi:10.1145/2623330.2623745

Cited by 77 publications

(2 citation statements)

References 28 publications

(66 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…To date, much research effort has been devoted to RWR, including its efficient computation ([6], [16], [5], [7], [18], [8], [14]), top- k search ([7], [10], [3], [17]), and various mining tasks underpinned by RWR ([13], [2], [12]). However, insufficient attention has been paid to a fundamental task that arises in many graph-related applications, which is to determine the source nodes that have a large amount of information flowing to a given query node.…”

Section: Introductionmentioning

confidence: 99%

“…First, methods ([6], [16], [14], [18]) have been proposed to compute the approximate RWR between any two nodes with an error bound ϵ . However, it is hard to pre-specify a proper ϵ for an ad-hoc query node q , because a pre-specified ϵ may be either too coarse to generate the correct top- k results, or too fine to avoid unnecessary computation.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Fast Inbound Top-K Query for Random Walk with Restart

Zhang

Jiang

Chen

et al. 2015

Machine Learning and Knowledge Discovery in Databases

View full text Add to dashboard Cite

Random walk with restart (RWR) is widely recognized as one of the most important node proximity measures for graphs, as it captures the holistic graph structure and is robust to noise in the graph. In this paper, we study a novel query based on the RWR measure, called the inbound top-k () query. Given a query node q and a number k, the query aims at retrieving k nodes in the graph that have the largest weighted RWR scores to q. queries can be highly useful for various applications such as traffic scheduling, disease treatment, and targeted advertising. Nevertheless, none of the existing RWR computation techniques can accurately and efficiently process the query in large graphs. We propose two algorithms, namely Squeeze and Ripple, both of which can accurately answer the query in a fast and incremental manner. To identify the top-k nodes, Squeeze iteratively performs matrix-vector multiplication and estimates the lower and upper bounds for all the nodes in the graph. Ripple employs a more aggressive strategy by only estimating the RWR scores for the nodes falling in the vicinity of q, the nodes outside the vicinity do not need to be evaluated because their RWR scores are propagated from the boundary of the vicinity and thus upper bounded. Ripple incrementally expands the vicinity until the top-k result set can be obtained. Our extensive experiments on real-life graph data sets show that queries can retrieve interesting results, and the proposed algorithms are orders of magnitude faster than state-of-the-art method.

show abstract

Section: Introductionmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Fast Inbound Top-K Query for Random Walk with Restart

Zhang

Jiang

Chen

et al. 2015

Machine Learning and Knowledge Discovery in Databases

View full text Add to dashboard Cite

show abstract

Parallel computations of local PageRank problem based on Graphics Processing Unit

Lai

Shao

et al. 2017

Concurrency and Computation

View full text Add to dashboard Cite

Summary In this paper, we study the issue of improving the performance of Markov chain Monte Carlo method to solve local PageRank problem under General Purpose Graphics Processing Unit environment. As a large number of dangling vertices cause large storage space of dangling vertices and thus slow down the Markov chain procession, we propose a reordering strategy to compress the storage space and reduce the computational complexity of Markov chain procession. In our performance study, by parallelizing and optimizing the proposed algorithm based on GPU, the reordering strategy can be up 12× faster compared with basic method, where the graphs have high‐proportion dangling vertices. According to our investigation on this issue, the variance of random walks determines the number of random walks in the computation; we thus introduce low‐discrepancy sequences to enhance the performance. Moreover, the low‐discrepancy sequences are organized to load in the on‐chip shared memory to accelerate fetching with a wise warp scheduling for bank conflict schema. A series of experiments have been conducted to evaluate the optimization efficiency. Compared with fetching data from off‐chip global memory, the shared‐memory‐based strategy can have over 10× speedup ratio performance. The experiments indicate that the size of shared memory has a significant impact on the parallelism of the proposed method as well.

show abstract