BLOKS solver scales Kernel Ridge Regression with novel preconditioner

Researchers have introduced BLOKS, a novel Conjugate Gradient-based solver designed for Nyström-sketched Kernel Ridge Regression (KRR). This method combines a block Jacobi approximation with sketching to create a preconditioner with memory costs independent of dataset size and model parameters, enabling scalability to large models. An efficient, GPU-accelerated implementation of BLOKS demonstrates comparable or superior performance to existing state-of-the-art methods on large-scale datasets, achieving excellent accuracy while maintaining favorable time and memory complexity. AI

IMPACT Introduces a scalable solver for kernel methods, potentially enabling larger and more complex models in machine learning research.

RANK_REASON Academic paper detailing a new algorithmic method for machine learning. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv stat.ML →

• alphaXiv
• arXiv
• block Jacobi approximation
• Bloks
• CatalyzeX Code Finder for Papers
• conjugate gradient method
• CORE Recommender
• DagsHub
• Gotit.pub
• graphics processing unit
• Hugging Face
• IArxiv Recommender
• Kernel Ridge Regression
• Nyström-sketched KRR
• ScienceCast

• paper
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →