Approximate Graph Propagation

Wang, Hanzhi; He, Mingguo; Wei, Zhewei; Wang, Sibo; Yuan, Ye; Du, Xiaoyong; Wen, Ji-Rong

doi:10.1145/3447548.3467243

Cited by 35 publications

(30 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The graph connectivity is represented by the adjacency matrix with self-loops as 𝑨 ∈ R 𝑛×𝑛 , while the diagonal degree matrix is 𝑫 ∈ R 𝑛×𝑛 . Following [8,28], we generalize the adjacency matrix of 𝐺 normalized by 𝑫 with convolution coefficient 𝑟 ∈ [0, 1] as Ã(𝑟) = 𝑫 𝑟 −1 𝑨𝑫 −𝑟 . For each node 𝑣 ∈ 𝑉 , denote the set of the out-neighbors of 𝑣 by N (𝑣) = {𝑢 |(𝑣, 𝑢) ∈ 𝐸}, and the out-degree of 𝑣 by 𝑑 (𝑣) = |N (𝑣)|.…”

Section: Preliminaries and Related Workmentioning

confidence: 99%

“…On a graph 𝐺, given a source node 𝑠 ∈ 𝑉 and a target node 𝑡 ∈ 𝑉 , the PPR 𝜋 (𝑠, 𝑡) represents the probability of a random walk with teleport factor 𝛼 ∈ (0, 1) which starts at node 𝑠 stops at 𝑡. In general, forward PPR algorithms, often categorized as single-source PPR, start the computation from 𝑠, comparing to backward or reverse alternatives that are developed from 𝑡 [28].…”

Section: Feature-pushmentioning

confidence: 99%

“…Denote a matrix 𝑿 = (𝒙 1 , • • • , 𝒙 𝐹 ), where 𝒙 𝑓 (1 ≤ 𝑓 ≤ 𝐹 ) is the 𝑓 -th column vector that is of length 𝑛 𝑈 and sum of elements is 1. Following [28], we assume all the entries 𝑥 𝑓 (𝑢) ≥ 0 for each 𝑢 ∈ 𝑈 . We use 𝜋 (𝒙; 𝑡) to represent the PPR for feature vector 𝒙, and can be defined as the probability of the event that a random walk which starts at a node 𝑠 ∈ 𝑈 with probability distribution 𝒙 and stops at 𝑡.…”

Section: Feature-pushmentioning

confidence: 99%

“…For Papers100M which has about 40GB raw data, even the 256GB capacity is insufficient for GBP to complete precomputation. In fact, we note that in [28], it utilizes up to 512GB RAM to conduct the experiment of GBP on Papers100M.…”

Section: Baseline Comparisonmentioning

confidence: 99%

“…Unfortunately, such methods are still not efficient enough when applied to million-scale or even larger graphs. According to [28], the very recent state-of-the-art algorithm GBP [8] still consumes more than 10 4 seconds solely for precomputation on the Papers100M graph (111M nodes, 1.6B edges, generated from MAG) to reach proper accuracy. In our experiments, the same model even exceeds the 160GB RAM bound on a single worker during learning.…”

Section: Introductionmentioning

confidence: 99%

See 4 more Smart Citations

SCARA: Scalable Graph Neural Networks with Feature-Oriented Optimization

Liao¹,

Mo²,

Luo³

et al. 2022

Preprint

View full text Add to dashboard Cite

Recent advances in data processing have stimulated the demand for learning graphs of very large scales. Graph Neural Networks (GNNs), being an emerging and powerful approach in solving graph learning tasks, are known to be difficult to scale up. Most scalable models apply node-based techniques in simplifying the expensive graph message-passing propagation procedure of GNN. However, we find such acceleration insufficient when applied to million-or even billion-scale graphs. In this work, we propose SCARA, a scalable GNN with feature-oriented optimization for graph computation. SCARA efficiently computes graph embedding from node features, and further selects and reuses feature computation results to reduce overhead. Theoretical analysis indicates that our model achieves sub-linear time complexity with a guaranteed precision in propagation process as well as GNN training and inference. We conduct extensive experiments on various datasets to evaluate the efficacy and efficiency of SCARA. Performance comparison with baselines shows that SCARA can reach up to 100× graph propagation acceleration than current stateof-the-art methods with fast convergence and comparable accuracy. Most notably, it is efficient to process precomputation on the largest available billion-scale GNN dataset Papers100M (111M nodes, 1.6B edges) in 100 seconds.

show abstract

Section: Preliminaries and Related Workmentioning

confidence: 99%

Section: Feature-pushmentioning

confidence: 99%

Section: Feature-pushmentioning

confidence: 99%

Section: Baseline Comparisonmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

SCARA: Scalable Graph Neural Networks with Feature-Oriented Optimization

Liao¹,

Mo²,

Luo³

et al. 2022

Preprint

View full text Add to dashboard Cite

show abstract

Scalable decoupling graph neural network with feature-oriented optimization

Liao,

Mo,

Luo

et al. 2023

The VLDB Journal

View full text Add to dashboard Cite

FICOM: an effective and scalable active learning framework for GNNs on semi-supervised node classification

Zhang,

Huang,

Zhang

et al. 2024

The VLDB Journal

View full text Add to dashboard Cite

Active learning for graph neural networks (GNNs) aims to select B nodes to label for the best possible GNN performance. Carefully selected labeled nodes can help improve GNN performance and hence motivates a line of research works. Unfortunately, existing methods still provide inferior GNN performance or cannot scale to large networks.Motivated by these limitations, in this paper, we present FICOM, an effective and scalable GNN active learning framework. Firstly, we formulate the node selection as an optimization problem where we consider the importance of a node from (i) the importance of a node during the feature propagation with a connection to the personalized PageRank (PPR), and (ii) the diversity of a node brings in the embedding space generated by feature propagation. We show that the defined problem is submodular, and a greedy solution can provide a $$(1-1/e)$$ ( 1 - 1 / e ) -approximate solution.However, a standard greedy solution requires getting the node with the maximum marginal gain of the objective score in each iteration, which incurs a prohibitive running cost and cannot scale to large datasets. As our main contribution, we present FICOM, an efficient and scalable solution that provides $$(1-1/e)$$ ( 1 - 1 / e ) -approximation guarantee and scales to graphs with millions of nodes on a single machine. The main idea is that we adaptively maintain the lower- and upper-bound of the marginal gain for each node v. In each iteration, we can first derive a small subset of candidate nodes and then compute the exact score for this subset of candidate nodes so that we can find the node with the maximum marginal gain efficiently. Extensive experiments on six benchmark datasets using four GNNs, including GCN, SGC, APPNP, and GCNII, show that our FICOM consistently outperforms existing active learning approaches on semi-supervised node classification tasks using different GNNs. Moreover, our solution can finish within 5 h on a million-node graph.

show abstract

Approximate Graph Propagation

Cited by 35 publications

References 31 publications

SCARA: Scalable Graph Neural Networks with Feature-Oriented Optimization

SCARA: Scalable Graph Neural Networks with Feature-Oriented Optimization

Scalable decoupling graph neural network with feature-oriented optimization

FICOM: an effective and scalable active learning framework for GNNs on semi-supervised node classification

Contact Info

Product

Resources

About