PHD: Compact Neural Network topology learning for on-sensor inference M/F

Job description

General information

Reference

2019-3034  

Job level

30 - Graduate Entry Level

Position description

Posting title

PHD: Compact Neural Network topology learning for on-sensor inference M/F

Regular/Temporary

Temporary

Job description

You will PhD student withing the Imaging division of STMicroelectronics in Grenoble with close collaboration with CEA-Leti, in the context of the Grenoble Multidisciplinary Institute in Artificial Intelligence (MIAI). Your work will be part of a collaborative project between Gipsa-Lab, Tima-Lab, ST and CEA-Leti, called “Near-Sensor Neural Computing”.

Current developments on image sensors tend to imply embedded computer vision in order to upgrade imager capabilities, in particular to enable machine learning based inference. In the case of embedded imaging systems, decision-making now becomes the core device feature. However, such a system still remains limited due to its computational and memory load, embedded algorithm complexity, dynamic power consumption and leakage. In that context, major challenges thus lie in power consumption and silicon footprint, which both rely on data bandwidth (i.e., signal dimensions and quantization levels).

Taking into account all the aforementioned issues, the objective of this PhD will be to define strategies for automatic selection -or learning [6]- of the best network topology under identified hardware constraints (memory requirements, complexity in terms of number of clock cycles, silicon surface and energy).

 

Profile

Education Level Required - BAC+5 (INGENIEUR,DESS,DEA...)

Desired Competencies are: Neural Network, Computer Visson

Others skills: python, data analysis

 

Position localisation

Job location

Europe, France, Grenoble

Candidate criteria

Education level required

5 - Master degree

Experience level required

Less than 2 years

Languages

English (2- Business fluent)

Requester

Desired start date

01/11/2019