Similarity Search with Tensor Core Units

2020/06/22 by Ahle, Thomas D., Silvestri, Francesco
#Data Structures and Algorithms (cs.DS) #Distributed #FOS: Computer and information sciences #Information Retrieval (cs.IR) #Parallel #and Cluster Computing (cs.DC)

paper · doi:10.48550/arxiv.2006.12608

Abstract

Tensor Core Units (TCUs) are hardware accelerators developed for deep neural networks, which efficiently support the multiplication of two dense √(m)× √(m) matrices, where m is a given hardware parameter. In this paper, we show that TCUs can speed up similarity search problems as well. We propose algorithms for the Johnson-Lindenstrauss dimensionality reduction and for similarity join that, by leveraging TCUs, achieve a √(m) speedup up with respect to traditional approaches.

Related

• A Computational Model for Tensor Core Units
• TCUDB: Accelerating Database with Tensor Processors
• Leveraging GPU Tensor Cores for Double Precision Euclidean Distance Calculations
• Fast and Scalable Mixed Precision Euclidean Distance Calculations Using GPU Tensor Cores
• Learning from distinctive candidates to optimize reduced-precision convolution program on tensor cores