Smart Light: Machine Learning control of lasers and applications in nonlinear optics
Machine learning is a field of artificial intelligence that applies
advanced techniques from statistics and numerical analysis to carry out
tasks without explicit programmed instructions. Recent years have seen
dramatic impact of machine learning in society, with commonplace
applications in health-care, autonomous vehicles, and language
processing. The impact of machine learning on basic research has been
just as significant, and the use of advanced algorithmic tools in data
analysis has resulted in new insights into many areas of science.
The objective of this thesis is to apply techniques of machine
learning to understand and exploit nonlinear propagation in optical
systems to develop customized and programmable light sources. In
particular, the aim will be to focus on ultrafast laser sources
producing picosecond and femtosecond pulses, and to develop deep
learning (neural network) approaches to both aid in the overall design
of the laser sources themselves, as well as to optimize the generation
and propagation of these pulses in nonlinear optical fibre. A parallel
numerical and theoretical objective is to use deep learning techniques
to determine underlying propagation equation models from analysis only
of experimental data.
Profile of the Candidate
It is expected that typical applicants will have a strong background
in physics, electrical engineering, or engineering science, with
previous training in optoelectronics, optics and laser physics an
advantage. The thesis can be oriented around either
theoretical/numerical work or experiment. For the latter, previous
experimental experience in nonlinear fibre optics is desirable, and in
both cases, applicants must have excellent computer skills and ideally
experience in Python and GPU Programming. Candidates with excellent
skills in MATLAB will also be very welcome to apply but it would be
expected that algorithmic tool development will require conversion to
Python. For focus on experimental work, experience with interfacing to
laboratory equipment will also be an asset, but the main requirement is a
passion and desire to learn advanced techniques at the state of the art
of artificial intelligence applied to one of the most exciting fields
of research in optics. In addition, given that a significant component
of the subject could be oriented towards algorithmic development of
machine learning methodologies, candidates from other fields such as
information science or applied mathematics are also encouraged to apply.
Contract Details
36 months starting from October 2020. Salary at the level fixed by
thesis funding from the French Ministry of Higher Education and Research
and Innovation. Approximately 1900 €/month before tax.
Work Environment
FEMTO-ST is one of France’s leading research centers in engineering
science and photonics, and combines permanent staff from the CNRS and
Université Bourgogne Franche-Comté in an exciting multidisciplinary
environment. Besançon is a picturesque town dating back to Roman times
containing a UNESCO World Heritage Site, many cultural attractions, and
easy access to outdoor pursuits.
Contact Information
Please contact Professor John Dudley john.dudley@univ-fcomte.fr