Advances in Efficient Graph Processing and Neural Network Techniques

The field of graph processing and neural networks is rapidly advancing, with a focus on developing efficient and scalable techniques for handling large-scale data. Recent developments have centered around improving the performance of graph neural networks, with innovations in mini-batching, similarity search, and substructure discovery. Notably, the use of disk-based similarity search and structure-aware randomized mini-batching has led to significant improvements in training time and accuracy. Additionally, the application of Gaussian processes to graph-based problems has shown promise, with low-rank computation methods enabling efficient posterior mean calculation. Some noteworthy papers in this area include:

  • Industrial-Scale Neural Network Clone Detection with Disk-Based Similarity Search, which demonstrates the effectiveness of disk-based similarity search for clone detection.
  • Efficient GNN Training Through Structure-Aware Randomized Mini-Batching, which presents a novel methodology for improving GNN training efficiency.
  • Efficient Learning on Large Graphs using a Densifying Regularity Lemma, which introduces a low-rank factorization of large directed graphs for efficient learning.

Sources

Industrial-Scale Neural Network Clone Detection with Disk-Based Similarity Search

Efficient GNN Training Through Structure-Aware Randomized Mini-Batching

Efficient Learning on Large Graphs using a Densifying Regularity Lemma

Subexponential and Parameterized Mixing Times of Glauber Dynamics on Independent Sets

A Taylor Series Approach to Correction of Input Errors in Gaussian Process Regression

Scalable Substructure Discovery Algorithm For Homogeneous Multilayer Networks

Graph-based Semi-supervised and Unsupervised Methods for Local Clustering

Efficient Graph-Based Approximate Nearest Neighbor Search Achieving: Low Latency Without Throughput Loss

Low-rank computation of the posterior mean in Multi-Output Gaussian Processes

S3AND: Efficient Subgraph Similarity Search Under Aggregated Neighbor Difference Semantics (Technical Report)

Built with on top of