Photonic co-processors in HPC: using LightOn OPUs for Randomized Numerical Linear Algebra
This addresses a key performance issue in HPC for large-scale linear algebra applications, though it is incremental as it applies an existing hardware method to a known bottleneck.
The paper tackles the computational bottleneck of randomization in Randomized Numerical Linear Algebra (RandNLA) for High Performance Computing by using LightOn Optical Processing Units to accelerate linear random projections, achieving significant speed-ups with negligible precision loss in algorithms like RandSVD and trace estimators.
Randomized Numerical Linear Algebra (RandNLA) is a powerful class of methods, widely used in High Performance Computing (HPC). RandNLA provides approximate solutions to linear algebra functions applied to large signals, at reduced computational costs. However, the randomization step for dimensionality reduction may itself become the computational bottleneck on traditional hardware. Leveraging near constant-time linear random projections delivered by LightOn Optical Processing Units we show that randomization can be significantly accelerated, at negligible precision loss, in a wide range of important RandNLA algorithms, such as RandSVD or trace estimators.