Richard Schoonhoven

Full Name
R.A. Schoonhoven
Function(s)
PhD student
Email
Richard.Schoonhoven@cwi.nl
Telephone
+31 20 592 4176
Room
L018
Department(s)
Computational Imaging

Biography

At university I obtained bachelors degrees in Mathematics and Physics at the university of Utrecht. Afterwards I obtained master degrees in Mathematics (focus on Algebraic Geometry) and Computer science (Focus on Algorithms and Complexity). Now, I am working on implementing and accelerating high-throughput pipelines for real-time tomography, and the CORTEX project. My work deals with deep learning elements, as well as optimization and graph theoretic algorithms. Our first project was to get real-time segmentation working in the CWI FleX-Ray lab. Next, I accelerated the deep learning elements (and several others) using a novel pruning method called LEAN. LEAN introduced the idea of scoring all the operators in a CNN with a mathematical operator norm, and extract the strongest paths in the resulting graph. On the hardware side I am working on tuning GPU kernels together with the eScience Centre. Here we are developing tailor-made GPU tuners (embedded in Kernel Tuner) with as final goal to accelerate the Amber pipeline in use at Astron.

Publications

Current projects with external funding

  • the Center for Optimal, Real-Time Machine Studies of the Explosive Universe (CORTEX)