Matthieu Simeoni  Matthieu Simeoni photo       

contact information

PhD Student
Zurich Research Laboratory, Zurich, Switzerland
  +41dash44dash724dash87dash61

links


more information

More information:  My LinkedIn profile  |  EPFL Personal Webpage

profile


Matthieu Simeoni originally joined the IBM Research – Zurich Lab as a student intern in 2014. He is currently a PhD student at the Swiss Federal Institute of Technology (EPFL) in Lausanne and a member of the Foundations of Cognitive Solutions research group under the supervision of Dr. Paul Hurley.

Matthieu received his M.Sc. degree in Engineering from EPFL in 2015 for his thesis entitled “A New Imager for the Square Kilometer Array.”

Among other projects, he and his group are responsible for the work package WP6 of the DOME project. This initiative, which involves researchers from both IBM Research and ASTRON (The Netherlands Institute for Radio Astronomy), is developing the technologies needed for next-generation radio interferometers such as the Square Kilometre Array (SKA), which, upon completion, will constitute the largest radio telescope ever built.

More specifically, the group of Dr. Paul Hurley is mainly interested in the elaboration and design of novel signal processing algorithms to optimize the data acquisition and processing chain in order to propose a processing pipeline adapted to the scale of the SKA.

Matthieu's own research work takes root in this multidisciplinary environment at the frontier between signal processing and applied mathematics. The main contribution of his work consists of a new imaging pipeline, which is both more accurate and faster than the current imaging pipeline. This new imager is particularly interesting for hierarchically designed interferometers, which reduce the data flow sent to the central processor by beamforming together the signals coming from groups of antennas.

Indeed, in this case, the classical Fourier relationship that links the data with the underlying sky image breaks down, and the unsatisfactory attempts to retain the Fourier framework have led to a computational overhead and a lot of non-intuitive algorithm tweaking. Hence, Matthieu described in his thesis how re-examining the problem from an orthogonal perspective (literally) results in a more intuitive, linear, and flexible chain.

The proposed imager works in two steps. First, a preconditioning based on the Gram–Schmidt orthogonalization procedure is performed in order to facilitate the computation of the pseudoinverse sky estimate. From this, the LASSO estimate is approximated very efficiently by means of thresholding. The quality of this approximation is shown to be linked to the effective support of the instrument point spread function. This algorithm is numerically stable even for signal-to-noise ratios as high as −24 dB. In the case of LOFAR (currently the world's largest radio telescope), we show that our algorithm can be 2 to 34 times faster than the state-of-the-art. Finally, the accuracy and sensitivity of the new technique are shown to be superior for simulated data.

Matthieu received the IBM Research Prize in Computational Science from EPFL in 2015.

Mathumaritarian logoIn addition to his studies and research activities, Matthieu co-founded the “Mathumanitarian Games” outreach project sponsored by EPFL, where students travel to South African townships to teach mathematics to school children in a fun and playful way. “Since the end of Apartheid, the education system in South Africa has suffered large disparities between urban and rural areas and between public and private schools,” Matthieu explains. “In the public schools, teachers are volunteers — not professionals. They have few resources, and classes are overcrowded.”