User Tools

Site Tools


en:projects

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Both sides previous revision Previous revision
en:projects [2019/03/11 17:08]
francoislp
en:projects [2019/04/08 16:33] (current)
francoislp
Line 9: Line 9:
  
 ===Using redundancy for energy efficiency in various contexts=== ===Using redundancy for energy efficiency in various contexts===
-  * [[en:​faultyLDPC|LDPC Decoders]]: The decoding of error-correction codes consumes an important portion of most telecommunication receivers, and they are also the perfect example of an algorithm that is naturally redundant. It is therefore very interesting to study how to save energy by running decoders on faulty hardware. In addition, ​I am interested in developing tools to design error-correction codes by starting from energy constraints. Having a precise control over decoding energy would allow to make better use of available resources in energy-constrained systems such as sensor nodes in an IoT context, and also to increase the use of powerful error-correction codes within integrated circuits. [[en:​faultyLDPC|(More details...)]]+  * [[en:​faultyLDPC|LDPC Decoders]]: The decoding of error-correction codes consumes an important portion of most telecommunication receivers, and they are also the perfect example of an algorithm that is naturally redundant. It is therefore very interesting to study how to save energy by running decoders on faulty hardware. In addition, ​we are interested in developing tools to design error-correction codes by starting from energy constraints. Having a precise control over decoding energy would allow to make better use of available resources in energy-constrained systems such as sensor nodes in an IoT context, and also to increase the use of powerful error-correction codes within integrated circuits. [[en:​faultyLDPC|(More details...)]]
  
   * [[en:​faultydnn|Deep neural networks]]: Algorithms based on deep neural networks achieve outstanding performance on a long list of difficult tasks such as image classification and segmentation,​ voice recognition,​ or playing the game of Go. However, to obtain these results, very large networks are required, containing several million parameters or more. As a result the energy required for the inference step makes it challenging to deploy such algorithms directly on embedded systems. Our research work seeks to dramatically reduce this energy consumption by building hardware accelerators that can function with unreliable circuits. [[en:​faultydnn|(More details...)]]   * [[en:​faultydnn|Deep neural networks]]: Algorithms based on deep neural networks achieve outstanding performance on a long list of difficult tasks such as image classification and segmentation,​ voice recognition,​ or playing the game of Go. However, to obtain these results, very large networks are required, containing several million parameters or more. As a result the energy required for the inference step makes it challenging to deploy such algorithms directly on embedded systems. Our research work seeks to dramatically reduce this energy consumption by building hardware accelerators that can function with unreliable circuits. [[en:​faultydnn|(More details...)]]
  
   * //​c//​-partite associative memories: This associative memory structure invented by Gripon and Berrou can [[http://​ieeexplore.ieee.org/​document/​7296693/​|be shown]] to be very robust to faulty hardware. It also has an efficient hardware implementation,​ making it attractive as a specialized cache memory, for instance for [[http://​ieeexplore.ieee.org/​document/​6560648/​|network routing]].   * //​c//​-partite associative memories: This associative memory structure invented by Gripon and Berrou can [[http://​ieeexplore.ieee.org/​document/​7296693/​|be shown]] to be very robust to faulty hardware. It also has an efficient hardware implementation,​ making it attractive as a specialized cache memory, for instance for [[http://​ieeexplore.ieee.org/​document/​6560648/​|network routing]].
en/projects.txt ยท Last modified: 2019/04/08 16:33 by francoislp