- Type
- Research
- Focus
- PCA · PyTorch · CUDA
- Context
- Astronomy and Computing
My contribution
Contributed software to qrpca, a QR-decomposition PCA package using PyTorch and CUDA acceleration.
Scientific computing
The package uses QR decomposition for principal component analysis and supports GPU acceleration. Published benchmarks reported 10–20× faster computation than conventional implementations on large matrices; this is a result reported for the package, rather than a universal speed-up for every workload.
Publication
R. S. de Souza, X. Quanfeng, S. Shen, C. Peng, and Z. Mu. qrpca: A Package for Fast Principal Component Analysis with GPU Acceleration. Astronomy and Computing, 2022.