This position is closed
## Open postdoctoral position:
## IntroductionThe research group of Prof. Laurent Jacques in the Image and Signal Processing Group (ISPGroup) of the University of Louvain-la-Neuve in Belgium (UCL) opens one position for a postdoctoral researcher to work on "ALTERSENSE" ("Computational Sensing Strategies for Low-Complexity Signal Models") a new project funded by the Belgian Fund for Scientific Research - FNRS. ## AlterSense ProjectWith the steady development of technology in numerous scientific fields such as biomedical sciences, astronomy, optics or computer vision, big challenges are raised by the design of new and efficient data acquisition systems. These must often comply with contradictory goals such as sampling high dimensional domains, devising fast and low-complexity recording processes, reaching low-power consumption sensors, facing limited capacity communication channels and being at the same time robust against multiple noise sources. Noticeably, the final objective of those sensors is invariably the same: providing at the very end of the data sensing chain processed and interpretable information, either for human or for automatic (machine) processing. This is the case, for instance, - in
**Satellite or Biomedical Imaging**: for numerous imaging technologies, such as Hyperspectral Imaging, Magnetic Resonance Imaging, Computed Tomography, or Positron-Electron Tomography,**segmenting**images or data volumes in a few categories of connected pixel areas (e.g., spectral endmembers, biological tissues) is of particular interest for simplifying information; - in
**Low-Power Dynamic Sensing**: in the general development of the Internet-of-Things (IoT) or of ultra-low energy sensors (e.g., biomedical).
As already formalized by Kolmogorov in the 60’s, all those applications are possible since “meaningful signals follow low-complexity descriptions”: their informative content is materialized by highly structured “patterns” whose intrinsic parameterization is considerably reduced compared to the high dimensionality of the ambient domain. By contrast, purely noisy signals often carry no information content, they are highly unstructured and require much more parameters to be characterized. Leveraging the paradigm shift introduced by the Compressed Sensing theory where signal sensing is adjusted to prior signal models, - for ubiquitous data processing tasks: for detecting, segmenting or classifying informative signals;
- for high-dimensional signals (e.g., hyperspectral or dynamic images) following low-complexity descriptions such as sparse/low-rank signal models or linear dynamical systems (LDS);
- for conveniently balancing sensing time/complexity, data quantization and transmission (as in 1-bit CS), final data processing accuracy and data processing time as any other limited resources.
## Research descriptionIn the general context described above, the postdoctoral researcher will more particularly focus his/her research during 2 years on one or several of the following topics: - classifications of low-complexity signals in high-dimensional space from their quantized and random projections, e.g., using 1-bit or universal quantization, establishing theoretical guarantees and numerical procedures for achieving this goal;
- design of fast quantized random embeddings leveraging structured sensing matrices with fast matrix-vector multiplication (e.g., Fourier, DCT, Hadamard), including a theoretical study of the connection relating the minimal embedding dimension, the intrinsic dimension of the embedded signal set and the accuracy (distortion) of the embedding;
- design of quantized random embedding for low-complexity set of signals in continuous domains (Hilbert space, manifold);
- provably convergent numerical methods for blind calibration of compressive sensing devices corrupted by, e.g., unknown gains, convolutions.
- combination of quantized random embedding with machine learning algorithms (SVM, K-means, ...) and study of the impact of the first on the performance of the second.
All those topics are considered as key tools for meeting the challenges raised by the AlterSense Project. ## Related publications
- L. Jacques, J. N. Laska, P. T. Boufounos, and R. G. Baraniuk, "Robust 1-Bit Compressive Sensing via Binary Stable Embeddings of Sparse Vectors", IEEE Transactions on Information Theory, Vol. 59(4), pp. 2082-2102, 2013. doi:10.1109/TIT.2012.2234823, arXiv:1104.3160 (submitted in April 2011)
- A. Moshtaghpour, L. Jacques, V. Cambareri, K. Degraux and C. De Vleeschouwer. "Consistent Basis Pursuit for Signal and Matrix Estimates in Quantized Compressed Sensing" IEEE Signal Processing Letters, vol. 23, no. 1, p. 25 - 29, january 2016. doi:10.1109/LSP.2015.2497543, arXiv:1507.08268
- L. Jacques, "A Quantized Johnson Lindenstrauss Lemma: The Finding of Buffon's Needle", IEEE Transactions in Information Theory, vol. 61, no. 9, p. 5012-5027 (Sept. 2015), doi:10.1109/TIT.2015.2453355, arXiv:1309.1507.
- L. Jacques, P. Vandergheynst, A. Bibet, V. Majidzadeh, A. Schmid, and Y. Leblebici, "CMOS Compressed Imaging by Random Convolution", Proc. of ICASSP'09, pp. 1113-1116, 19-24 April 2009, Taipei, Taiwan. doi:10.1109/ICASSP.2009.4959783, preprint
- L. Jacques, "Small width, low distortions: quasi-isometric embeddings with quantized sub-Gaussian random projections", Submitted, arXiv:1504.06170
- P. T. Boufounos, L. Jacques, F. Krahmer, and R. Saab, "Quantization and Compressive Sensing" Chapter of the book "Compressed Sensing and Its Applications", Edited by H. Boche, R. Calderbank, G. Kutyniok, J. Vybiral (Birkhäuser Mathematics, Springer 2015). ISBN:978-3-319-16041-2 (in combination with the MATHEON workshop). arXiv:1405.1194, doi:10.1007/978-3-319-16042-9_7
- L. Jacques and P. Vandergheynst, "Compressed Sensing: When sparsity meets sampling", In "Optical and Digital Image Processing - Fundamentals and Applications", Edited by G. Cristòbal; P. Schelkens; H. Thienpont. Wiley-VCH, April 2011. ISBN:978-3-527-40956-3, dial:2078.1/87547, doi:10.1002/9783527635245.ch23, preprint
## Applicant's profile:- PhD in applied mathematics, electrical engineering or theoretical physics
- Strong background in signal processing, compressed sensing and inverse problem solving.
- Knowledge in at least one of the following topics is a plus: measure concentration phenomenon, machine learning, signal detection, classification methods, high dimensional data processing, convex optimisation.
- Excellent programming skills in a numerical language (matlab or python);
- Good communications skills, both written and oral, in English.
- Knowing French *is not* required (the research group is international)
*Mobility criterion:*The position is open to all nationalities, including Belgian, but the condition is**to have spent less than 12 months over the last three years in Belgium**.
More information about the project can be obtained upon request. ## We offer:- A research position in a dynamic environment, working on leading-edge theories and applications with international contacts;
- A research team constituted of one professor, another postdoctoral researcher, and 3 PhD students on topics related to AlterSense;
- A 24-month position funded by the Belgian NSF (FNRS)
- The funding is a scholarship and Visa will be needed for a non-EU researcher.
## Application:Applications should include: - a detailed resume (in pdf) + list of publications;
- 2-page research statement (in pdf)
**explaining also why the candidate is interested in working in the research topics described above**and**how it is connected to his/her PhD background** - the names and complete addresses of two reference persons that can be contacted
Please send applications by email to: - Prof. Laurent Jacques, laurent.jacquesuclouvain.be
## Candidate Selection- Pre-selection of candidates based on their application files
- (remote) Interview of the short-listed candidates
- The successful candidate can be hired from June 1st 2016, by we have some flexibility to start the position (a bit) later (to be discussed)
Last updated June 27, 2017, at 10:33 AM |