GShuttle: Optimizing Memory Access Efficiency for Graph Convolutional Neural Network Accelerators

• Published in: Journal of computer science and technology Vol. 38; no. 1; pp. 115 - 127
• Main Authors: Li, Jia-Jun, Wang, Ke, Zheng, Hao, Louri, Ahmed
• Format: Journal Article
• Language: English
• Published: Singapore Springer Nature Singapore 01.02.2023; Springer; Springer Nature B.V
• Subjects: Accelerators; Analysis; Artificial Intelligence; Artificial neural networks; Computer Science; Data Structures and Information Theory; Design analysis; Design and construction; Design optimization; Dynamic random access memory; Efficiency; Energy efficiency; Graph neural networks; Information Systems Applications (incl.Internet); Mathematical optimization; Memory (Computers); Methods; Neural networks; Optimization techniques; Random sampling; Random variables; Regular Paper; Software Engineering; Theory of Computation; graph convolutional neural network; memory access; neural network accelerator

Graph convolutional neural networks (GCNs) have emerged as an effective approach to extending deep learning for graph data analytics, but they are computationally challenging given the irregular graphs and the large number of nodes in a graph. GCNs involve chain sparse-dense matrix multiplications w...