Andrew Thompson's homepage
Visiting Researcher, Mathematical Institute, University of Oxford
About me
I am a Senior Research Scientist at the National Physical Laboratory and a Visiting Researcher in the Mathematical Institute at the University of Oxford, and my research interests are in sparse estimation, compressed sensing and matrix completion, including applications to 5G wireless communications and image processing.

A new algorithm for the Walsh-Hadamard Transform
This short paper proposes a new algorithm for computing the Walsh-Hadamard Transform which consists entirely of Haar wavelet transforms. The approach is quite different from the familiar divide-and-conquer algorithms, and it appears to offer some intriguing advantages from the point of view of circuit design and parallelization.
Posted 19-5-2017 · paper · slides
Quantitative recovery guarantees for IHT algorithms
Photo One A precise asymptotic phase transition is known for recovering a sparse vector from incomplete Gaussian measurements using l_1 minimization. Many other recovery algorithms have now been proposed for this problem, such as Iterative Hard Thresholding (IHT). In this paper we prove the best known theoretical phase transitions for IHT and Gaussian matrices.
Posted 6-11-2014 · paper · slides
Optimal subspace packings
In this paper we provide deterministic constructions of optimal subspace packings with respect to spectral distance. We also show that random subspaces give close to optimal packings asymptotically. Our results can also be interpreted as matrix constructions with optimal block coherence.
Posted 6-11-2014 · paper · slides
Mathematical Institute
University of Oxford
Andrew Wiles Building
Woodstock Road
OX2 6GG.
phone: 01865 273579