Bringing UMAP Closer to the Speed of Light with GPU Acceleration

TL;DR

This paper introduces a GPU-accelerated version of the UMAP algorithm that achieves up to 100x speedup, making it more efficient and faithful, with broad applicability to other graph and manifold learning algorithms.

Contribution

The authors develop a set of techniques to significantly accelerate UMAP on GPUs, improving speed and accuracy over previous GPU implementations.

Findings

Achieved up to 100x speedup in practice

Developed techniques improving UMAP's fidelity on GPU

Implementation available in RAPIDS cuML library

Abstract

The Uniform Manifold Approximation and Projection (UMAP) algorithm has become widely popular for its ease of use, quality of results, and support for exploratory, unsupervised, supervised, and semi-supervised learning. While many algorithms can be ported to a GPU in a simple and direct fashion, such efforts have resulted in inefficient and inaccurate versions of UMAP. We show a number of techniques that can be used to make a faster and more faithful GPU version of UMAP, and obtain speedups of up to 100x in practice. Many of these design choices/lessons are general purpose and may inform the conversion of other graph and manifold learning algorithms to use GPUs. Our implementation has been made publicly available as part of the open source RAPIDS cuML library (https://github.com/rapidsai/cuml).