Advanced reinforcement learning for efficient optimization of analogue NNs

Updated: 2 months ago
Location: Besan on, FRANCHE COMTE
Job Type: FullTime
Deadline: 30 May 2025

15 Feb 2025
Job Information
Organisation/Company

CNRS-FEMTO-ST institute
Department

Optics
Research Field

Physics » Optics
Computer science » Programming
Researcher Profile

First Stage Researcher (R1)
Positions

PhD Positions
Country

France
Application Deadline

30 May 2025 - 12:00 (Europe/Paris)
Type of Contract

Temporary
Job Status

Full-time
Hours Per Week

38
Offer Starting Date

1 Jul 2025
Is the job funded through the EU Research Framework Programme?

Horizon Europe - MSCA
Is the Job related to staff position within a Research Infrastructure?

No

Offer Description

The candidate will be enrolled for PhD at the University Marie et Luis Pasteur, Besancon, at the CNRS laboratory FEMTO-ST and will work on leveraging advanced reinforcement and evolutionary learning techniques (e.g. co-variance matrix, policy gradient optimization) for training analogue NNs.
Secondments at CSIC and HPE will train the DC in neuromorphic approaches, in the industrial perspective on novel computing paradigms and in nonlinear photonic effects.
Efficiently training neuro-photonic systems currently is among the most important research directions. The doctoral candidate will explore reinforcement and evolutionary learning to improve the efficiency and performance of training analogue NNs. The crucial challenge that DC addresses is that probing each neuron and connection is energetically and hardware-complexity prohibitive, hence banning techniques like error back propagation or digital twin optimization. The candidate will study the convergence of different methods against their energy consumption using our demonstrated large area
vertical cavity surface emitting lasers (LA-VCSEL) NN as robust and demonstrated hardware platform.


Where to apply
E-mail

daniel.brunner@femto-st.fr

Requirements
Research Field
Technology » Computer technology
Education Level
Master Degree or equivalent

Research Field
Physics » Optics
Education Level
Master Degree or equivalent

Skills/Qualifications

High expertise level in neural network and machine learning coding using pytorch. Some experience with physics experiments would be helpful but is not mandatory.


Languages
ENGLISH

Additional Information
Work Location(s)
Number of offers available
1
Company/Institute
CNRS
Country
France
State/Province
Doubs
City
Besancon
Postal Code
25030
Street
15B Avenue des Montboucons
Geofield


Contact
City

Besancon
Website

http://www.femto-st.fr/en/
https://members.femto-st.fr/daniel-brunner/
Street

15B avenue des montboucons
Postal Code

25030
E-Mail

daniel.brunner@femto-st.fr

STATUS: EXPIRED

  • X (formerly Twitter)
  • Facebook
  • LinkedIn
  • Whatsapp

  • More share options
    • E-mail
    • Pocket
    • Viadeo
    • Gmail
    • Weibo
    • Blogger
    • Qzone
    • YahooMail