NOTE! This site uses cookies and similar technologies.

If you not change browser settings, you agree to it. Learn more

I understand

Learn more about cookies at :

October 22, 2018 | Easy reconfiguration of programmable components

raw 250
© zdyma4 -

Researchers at List*, a CEA Tech institute, won first prize in the RAW competition for its smart positioning and routing software for reconfigurable architectures.

The IEEE International Parallel and Distributed Processing Symposium 2018 coincided with the 25th anniversary of the Reconfigurable Architectures Workshop (RAW). To mark the milestone, a floor planning competition was held, where competitors were asked to develop algorithms to predict the positions of various elements of FPGA-type programmable components. List won first prize in the competition.

The challenge was to find the solution capable of reducing the surface area occupied by functional blocks to the bare minimum while reducing the total connection length as much as possible. List's algorithm used simulated annealing, a method for resolving optimization problems. The method keeps the program from getting "stuck" on a localized optimal configuration. However, to ensure that the best possible solution was generated, List developed a "multistart" approach that restarts the calculations with several different initial solutions to vary the results and find the optimal solution or get as close to the optimal solution as possible.

The program, which verifies all of the constraints of the problem (and others), made List the only competitor to obtain a solution for all 30 problems assigned for the competition. The win highlights List's know-how in operational research and combinatorial optimization applied to electronic circuit design automation.

*List earned the prestigious Institut Carnot seal in 2006 (Institut Carnot TN@UPSaclay).




October 16, 2018 | Shaping the vehicle of the future

List place au vhicule du futur
© zapp2photo- Fotolia

Tomorrow’s land transportation will be clean, smart, and shared. CEA Tech has been driving advances in the vehicle of the future through targeted research programs for years.

Researchers in the field of land transportation are tackling the challenges of making transportation clean, environmentally-friendly, and shared head-on. The CEA Tech institutes conduct research in a very wide range of land transportation technologies and, in particular technologies designed to address precisely these challenges. The institutes' programs encompass vehicles (cars, trains, etc.), infrastructures, and the associated services (car sharing, ride sharing, and multi-modal transportation systems), but that's not all. CEA Tech also conducts research that will help make vehicle manufacturing more reliable and efficient in the Factory of the Future.

CEA Tech labs are home to research projects on vehicle electrification (enhanced magnets for electric motors), batteries, hybrid combustion-electric systems, fuel cells, and hydrogen tanks. CEA Tech infrastructure-related research includes projects on fast charging terminals, optimized multi-user vehicle charging technologies compatible with smart grids, two-way (vehicle-to-grid) charging, and more.

Since 2014, CEA Tech has either coordinated or contributed to some 20 EU Horizon 2020 research and development projects targeting road transportation. CEA Tech has also created partnerships with some of the industry's major stakeholders, including OEM Renault and Tier-1 automotive suppliers like Valeo, which integrated the CEA's Deep Manta technology—an algorithm that enables driverless cars to recognize other cars in their vicinity—into the demo car it exhibited at CES Las Vegas in 2018, and Faurecia, which is working with CEA Tech on a fuel-cell technology.




October 11, 2018 | Biomargin gets biomarkers talking

biomargin marqueurs
©santypan –

​In research conducted under the Biomargin project, a series of reliable, robust biological signatures were identified to predict the risk of rejection after a kidney transplant. The biomarkers will help doctors make the best possible decisions for their patients.

A transplant is the leading treatment for terminal-stage kidney failure. It could soon be possible to predict a patient's risk of rejecting the transplant before surgery by analyzing certain biomarkers. The EU Biomargin project, which was completed in 2018, was set up with the goal of identifying and testing biomarkers to assess the risk of the three most common types of rejection: humoral immunity, cellular immunity, and fibrosis and atrophy of the transplanted organ's blood vessels.

List*, a CEA Tech institute, was one of the thirteen partners involved in the project. List researchers analyzed several types of biological data (messenger RNA, proteins, metabolites, and lipids) generated using several measurement techniques on biopsy, urine, and blood samples taken from a group of patients. Based on the analysis, around ten markers of interest were selected from among several thousand and original biological signatures were identified to predict the risk of rejection using statistical models developed specifically for this purpose.

The assessment of the multiomic[1] biomarkers' capacity to predict rejection revealed better performance than predictions made based on isolated biomarkers. The approach will now be tested on a larger cohort of patients and will perhaps ultimately be used by doctors to make better-informed decisions.

[1]based on different kinds of measurements (transcriptome, proteome, metabolome, lipidome, etc.)

*List earned the prestigious Institut Carnot seal in 2006 (Institut Carnot TN@UPSaclay).




October 1, 2018 | Faster, more energy efficient neural networks with NeuroSpike

neurospike list 250
© Sergey Tarasov –

List, a CEA Tech institute, successfully completed the optimized execution of a complete convolutional neural network impulse model for the first time ever using its patented NeuroSpike calculator.

Convolutional neural networks (CNN) are widely used in the latest 2D data processing methods—and, especially, image recognition and classification. List recently developed a calculator that makes it possible to execute CNNs optimally based on impulse models[1], which, theoretically, can reduce the hardware resources and energy required to process data (the “inference” step).

The use of impulse models for CNNs requires coding the data as impulses that are transmitted between a wide range of artificial neurons arranged in different types of layers. To get the most out of the data coded in this way and complete the calculations as efficiently as possible, a special hardware architecture is also required. List* developed its NeuroSpike calculator for this purpose. It is the first hardware architecture that enables the efficient calculation of a complete CNN’s impulse inference.

NeuroSpike uses eleven times less energy and is four times faster than impulse-based architectures used to calculate 2D convolution inferences. It is protected by three patents.

[1] Impulse models are a way of processing data coded as impulses transmitted within a neural network. The artificial impulse neurons use less complex and more energy efficient operators than traditional models, but they require more memory.

*List earned the prestigious Institut Carnot seal in 2006 (Institut Carnot TN@UPSaclay).




September 24, 2018 | DOSEO expands training for medical physicists

doseo formation List
© Godart

The DOSEO radiotherapy research and training center recently created a new training course combining theoretical and hands-on learning over three two-day sessions. The new course is being offered in response to high demand from medical physicists.

DOSEO’s mission is to ensure that new radiotherapy technologies are used optimally to guarantee maximum patient safety. The center also trains radiotherapy students and practitioners. DOSEO recently presented its latest training course to 70 attendees at Journée des Manipulateurs Radio 2018, an event for radiotherapy professionals.

DOSEO already offers a course in metrology for radiotherapy that draws on the center’s broad, deep knowledge of metrology. The new course, for medical physicists, was created at the request of the French national nuclear safety board (ASN). It focuses on measuring doses in stereotaxic conditions or, in other words, in minibeam therapy. So far the course has been so popular that several additional dates have been scheduled.

A third training course will provide physicists with an introduction to dose measurement in imaging, which requires specific protocols. The topic was the subject of the AID-IGRT project funded by the French national research agency (ANR), which set out to control the doses received by patients in image-guided radiotherapy. 3D calculation software was developed to measure the doses delivered by the embedded imaging systems, generating a personalized calculation of the dose received by each of the patient’s organs depending on the patient’s morphology and the unique parameters of the treatment protocol.


Philippe Giraud, President of AFCOR (the Association for Continuing Education in Oncology and Radiotherapy) and member of the DOSEO Steering Committee:

“DOSEO is the only technical facility that has state-of-the-art equipment and high-level experts on the same premises available to practitioners, dosimetry technicians, and medical physicists for training on the latest radiobiology, imaging, and medical physics techniques. And, because DOSEO is near INSTN (France’s national nuclear science and technology institute), it has access to vital medical physics know-how. Beyond training, DOSEO is also an active participant in research projects, helping to transfer the results of the projects to industry with Aviesan, France’s national alliance for the life and health sciences. It is our main resource for medical physics and radioprotection.”

 DOSEO website: