Hardworking Energy-efficient Learning Platform

About the Project

Deep learning is revolutionizing many areas but its resource needs are often disproportionate and limit its scope to projects without energy constraints and with few deployment constraints.

In the medical field for example, where the circulation of data is not free and where the means of calculation are limited, the use of advanced techniques derived from deep learning is often not possible.

The aim of the HELP project is the acquisition and implementation of a very high performance and low power consumption (based on a matrix of GPUs and low-power processors) dedicated to learning algorithms and in particular to deep neural networks.

The HELP platform will enable the design and optimization of parallel computing algorithms on deep neural networks, to evaluate the performance achievable in practice on such material and validate the feasibility of a deployment in a medical setting.

The HELP project  is fully transdociplinary and is be backed by the Idex project "Integration and Analysis of Biomedical Data". It will provide additional material and human resources for the deployment of a platform benefiting from the latest technological advances in low computing consumption. It will benefit from the work of data collection and development algorithmic realized in IADB.

Principal Investigator
Project's partner(s)
  • Stéphanes DESCOMBES, MSI
  • Fabien GANDON, Inria
  • Damon MAYAFFE, BCL

October 2017- March 2019

Total Amount

35,4 k€ (co-funding from the Academy of Excellence Human, Societies, Ideas and Environments)

Related Documentation