Linear algebra kernels and algorithms are prevalent in scientific computing, machine learning, statistics, image processing, and much more. With the ever-increasing scale and complexity of some of those computational problems, the need for scalable and cost-effective linear algebra algorithms is critical. For this reason, approximate computing techniques have been increasingly used in modern linear algebra algorithms. These techniques can substantially reduce the algorithms’ complexities by introducing (controlled) errors into the calculations. Those techniques include, but are not limited to, low-rank approximations, mixed precision, randomization, or tensor compression.

The Chinese Academy of Sciences Workshop on Approximate Computing in Numerical Linear Algebra (CAS-ANLA) aims to provide greater exposure to the topic of approximate computing techniques to the local Chinese and Asian scientific community. It will gather outstanding researchers and students from China working in approximate computing together with some of the best international experts in the field. The speakers will illustrate theoretically and experimentally the benefit of cutting-edge approximate computing techniques and present recent advances in the field. We also expect other topics to be represented, such as optimization, machine learning, or functions of matrices to name a few.

The event will take place in the South building of the Academy of Mathematics and Systems Science on April 23rd and 24th, 2025.

How to participate?

We are looking for speakers!

If you wish to participate or have a student who wishes to, please contact me at [email protected]. Note that the presentation has to be delivered in English. The format is a plenary talk, except for students and postdocs for which the format is 20 minutes of talk plus 5 minutes of questions.

Preliminary list of speakers