CUR Decomposition: A low-rank approximation with column/row subsets
An intuitive introduction to CUR approximation using actual matrix columns and rows, its practical advantages, and tips for constructing and stabilizing it.
Notes and exposition
Blog
Writing on numerical analysis, approximation theory, randomized numerical linear algebra, operator learning and related mathematics.
Latest writing
RSSColumn Subset Selection: how to choose a small representative subset from millions of data points?
Why a small subset of actual matrix columns can provide an interpretable, structure-preserving low-rank approximation, and how common CSSP algorithms choose them.
How can polynomial interpolation remove noise? (Part I)
A simple route from noisy Chebyshev interpolation to a fast, stable polynomial approximation using coefficient truncation and Mallows’ criterion.