Accelerating Barnes-Hut t-SNE Algorithm by Efficient Parallelization on Multi-Core CPUs
Machine Learning
2022-12-23 v1 Artificial Intelligence
Distributed, Parallel, and Cluster Computing
Abstract
t-SNE remains one of the most popular embedding techniques for visualizing high-dimensional data. Most standard packages of t-SNE, such as scikit-learn, use the Barnes-Hut t-SNE (BH t-SNE) algorithm for large datasets. However, existing CPU implementations of this algorithm are inefficient. In this work, we accelerate the BH t-SNE on CPUs via cache optimizations, SIMD, parallelizing sequential steps, and improving parallelization of multithreaded steps. Our implementation (Acc-t-SNE) is up to 261x and 4x faster than scikit-learn and the state-of-the-art BH t-SNE implementation from daal4py, respectively, on a 32-core Intel(R) Icelake cloud instance.
Keywords
Cite
@article{arxiv.2212.11506,
title = {Accelerating Barnes-Hut t-SNE Algorithm by Efficient Parallelization on Multi-Core CPUs},
author = {Narendra Chaudhary and Alexander Pivovar and Pavel Yakovlev and Andrey Gorshkov and Sanchit Misra},
journal= {arXiv preprint arXiv:2212.11506},
year = {2022}
}