Post-doc position
Software programming environment for neuromorphic accelerator
Area: Computer Science - Compilation..
The candidate should hold a Ph.D. in computer science. Expertise in either optimization algorithms, or FPGA place & route algorithms is particularly appreciated.
This post-doc position is part of the Arch²Neu project, funded by the French National Research Agency. This project aims at investigating the potential of neuromorphic architectures for computing purposes, and particularly for signal-processing applications. We are developing both the analog neuron hardware and the associated digital communications infrastructure, libraries of operators, and a compilation toolchain to provide the user with a robust, efficient, and easy-to-program accelerator for signal processing.
Signal processing tasks are decomposed into primitives (operators), that are well suited to spiking neural networks. Elementary operators can be composed to obtain larger applications, eventually mapped to the neural network hardware.
There is a need for an automated toolchain that maps a high-level application description onto the hardware. The candidate will be in charge of designing this compilation toolchain, and the associated software optimizations.
The candidate will strongly interact with the other members of the project-team working on architecture and library design.
Please contact
Rodolphe Héliot
Office: +33 4 38 78 17 69 -
email: rodolphe (dot) heliot (at) cea (dot) fr
www-leti.cea.fr
Olivier Temam
Office: +33 1 72 92 59 52 -
email: olivier (dot) temam (at) inria (dot) fr
www.inria.fr
Post-doc position
Computational abilities of spiking neurons-based computing
Area: Applied mathematics - signal processing.
The candidate should hold a Ph.D. in applied mathematics, electrical engineering, or computer science.
This post-doc position is part of the Arch²Neu project, funded by the French National Research Agency. This project aims at investigating the potential of neuromorphic architectures for computing purposes, and particularly for signal-processing applications. We are developing both the analog neuron hardware and the associated digital communications infrastructure, libraries of operators, and a compilation toochain to provide the user with a robust, efficient, and easy-to-program accelerator for signal processing.
A broad variety of signal-processing operators can be built using spiking neurons. These operators can then be composed to build signal processing tasks. In that perspective, it is vital to identify and characterize the classes of operators -and applications- than can be implemented using networks of spiking neurons. From the analytical model of the spiking neuron, the candidate will derive high-level properties of spiking neural networks, and contribute to the development of new operators and applications.
The candidate will tightly interact with the other members of the project-team working on architecture design and programming environment.
Please contact
Rodolphe Héliot
Office: +33 4 38 78 17 69 -
email: rodolphe (dot) heliot (at) cea (dot) fr
www-leti.cea.fr
Olivier Temam
Office: +33 1 72 92 59 52 -
email: olivier (dot) temam (at) inria (dot) fr
www.inria.fr
