Projects per year
Abstract
Distributed Stochastic Neighbor Embedding (tSNE) is a wellknown dimensionality reduction technique used for the visualization of highdimensional data. However, despite several improvements, tSNE is not wellsuited to handle large datasets. Indeed, for large datasets, the computation time required to obtain the visualizations is still too high to incorporate it in an interactive data exploration process. Since tSNE can be seen as an Nbody problem in physics, we present a new variant of tSNE based on a popular algorithm used to solve the Nbody problem in physics called ParticleMesh (PM). The problem is solved by first computing a potential in space and deriving from it the force exerted on each body. As the potential can be computed efficiently using Fast Fourier Transforms (FFTs), this leads to a significant speed up. The mathematical correspondence between tSNE and PM presented in this work could also lead to other future improvements since more advanced PM algorithms have been developed in physics for decades.
Original language  English 

Title of host publication  International Joint Conference on Neural Networks 
Publisher  IEEE 
Number of pages  8 
DOIs  
Publication status  Published  2021 
Publication series
Name  2021 International Joint Conference on Neural Networks (IJCNN) 

Keywords
 machine learning
 numerical physics
 dimensionality reduction
 visualization
 tSNE
 particleMESH
Fingerprint
Dive into the research topics of 'Accelerating tSNE using Fast Fourier Transforms and the ParticleMesh Algorithm from Physics'. Together they form a unique fingerprint.Projects
 1 Active

CÉCI – Consortium of high performance computing centers
CHAMPAGNE, B., Lazzaroni, R., Geuzaine , C., Chatelain, P. & Knaepen, B.
1/01/18 → 31/12/22
Project: Research
Equipment

High Performance Computing Technology Platform
Benoît Champagne (Manager)
Technological Platform High Performance ComputingFacility/equipment: Technological Platform