Numerical linear algebra has a broad and lively research frontier. New applications demand ever-greater scalability and accuracy, and evolutions in computer architecture require fundamentally new approaches to optimize performance on current and future machines. Prompted by these challenges, experts in the field are exploring new ways to use matrix structure, algorithms with efficient computation and communication patterns, and methods to automatically optimize implementations for high performance. Research in the area continues to have impact that is widely felt across many areas of science and engineering, largely because of a tradition of robust software design and exposition for broad audience.

This workshop is dedicated to James Demmel for his leadership in modern high-accuracy and high-performance linear algebra solvers and for his promotion of numerical linear algebra to a broad audience through software, teaching, and collaboration. Workshop themes include:

• Linear algebra software design
• Scalable communication-avoiding methods
• High-performance and auto-tuning
• High-accuracy computations and error analysis
• Application-inspired methods
• Connections between numerical linear algebra and other areas of mathematics
• Promotion of numerical linear algebra through teaching and exposition

Organizing committee

• Zhaojun Bai, UC Davis
• David Bindel, Cornell
• Inderjit Dhillon, UT Austin
• Ming Gu, UC Berkeley
• Xiaoye Li, LBNL

Sponsors