A new paper details significant speedups for Improved Kernel Partial Least Squares (IKPLS) algorithms, which are known for their efficiency in PLS calibration. The research introduces optimizations for computing X rotations and Y loadings, leading to parallelization improvements on modern hardware. Benchmarks using NumPy on CPUs and JAX on GPUs demonstrate speedups of up to two orders of magnitude for specific computational steps and approximately 2x to 6x for complete fits. These enhancements are implemented in the open-source Python package `ikpls`. AI
IMPACT Offers potential for faster data analysis in machine learning workflows.
RANK_REASON The cluster contains a research paper detailing algorithmic improvements and benchmarks. [lever_c_demoted from research: ic=1 ai=1.0]
- central processing unit
- graphics processing unit
- IKPLS
- Improved Kernel Partial Least Squares
- JAX
- NumPy
- Ole-Christian Galbo Engstrøm
- Python
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
