Kent-Lille Cotutelle: Novel neuromorphic, radically energy efficient training algorithms for action recognition

University of Kent
March 24, 2023
Offerd Salary:N/A
Working address:N/A
Contract Type:N/A
Working Time:N/A
Working type:N/A
Ref info:N/A
Scholarship value

Tuition fees and stipend at the standard Research Council rate (Home rate only: £4,596 (fees) and £17,668 (stipend) in 2022/23). The 2023/2024 rate is yet to be announced by the UK Research Councils.


Applications must be received by Sunday Friday 24 March 2023, 23:59 GMT


Open to home and international fee paying students. Home fees only are provided, the shortfall in international fees would need to be self-funded.

Scholarships are available on a cotutelle (dual award) basis only.

Students have to spend at least 12 months at Kent and Lille.

Key attributes and skills for prospective applicants:

To be successful, the applicant needs to have a first class undergraduate or master-level degree in computer science or a related subject (Statistics, Electrical engineering, etc.), ideally with a specialisation in machine learning.

Experience with one or more of the following is desirable (but not mandatory):

  • spiking neural networks,
  • artificial neural networks / deep learning,
  • computer vision,
  • neuromorphic hardware.
  • While this is not a software engineering project, strong skills in programming will be required, especially in Python.

    The applicant must also demonstrate scientific curiosity, the ability to work on both theory and practice, and good writing skills.

    Moreover, the student should be open to work with two different supervisors and should be comfortable to seek and take advice from the entire supervisory team, including handling possibly conflicting advice. Knowledge of French is desirable, but the willingness to learn is essential.

    Further details

    Kent-Lille Cotutelle: Novel neuromorphic, radically energy efficient training algorithms for action recognition.


  • Dr Dominique Chu, School of Computing, University of Kent
  • Professor Pierre Tirilly, Faculté des sciences et technologies, University of Lille
  • This project will develop novel algorithms for Spiking Neural Networks to detect gestures and actions in videos. The student will be jointly supervised by Prof. Pierre Tirilly at the Cristal Laboratory of the University of Lille and Dr. Dominique Chu at the School of Computing, University of Kent. The successful applicant will spend at least one year in Lille. The degree will lead to a dual doctoral award from Kent and from the University of Lille.

    Lille is one of the largest cities in France. It is located in northern France, close to the Belgian border, and connected by train to Paris (1h00), London (1b0), and Brussels (30 min.). In the University of Lille, the student will join the CRIStAL lab, a computer science and automatic control research center gathering over 450 researchers, faculty members, and PhD students, and IRCICA, an interdisciplinary research structure.

    Project specifics:

    Computing-related activities account now for a double digit percentage of the total global electricity consumption. Artificial Intelligence (AI) and especially the training of Neural Networks (NN) are known to be highly energy intense. In its current form, AI is thus environmentally unsustainable. At the same time, AI in general and NNs in particular have become societally important technologies and will also be important tools to find solutions to large-scale problems such as climate change. There is thus an urgent need to find ways to reduce the energy consumption of AI algorithms.

    A possible solution to the energy consumption of NNs is to use neuromorphic hardware to train them. Such hardware is radically more energy efficient than general purpose computers and GPUs. However, it requires a variant of NNs, so- called spiking neural networks (SNN), which are less researched than standard NNs. Moreover, the backpropagation algorithm cannot be run on this hardware. There are some alternatives that are compatible with neuromorphic hardware, but these currently underperform relative to backpropagation. For this project, we will develop novel algorithms to train SNNs on neuromorphic hardware to achieve performances that are comparable to backpropagation.

    This PhD project will focus on a particular AI problem: Automated recognition of actions or gestures performed by humans or animals in video clips. Typical solutions to this problem are based on deep neural networks trained with backpropagation that extract motion features from the video clips to help identify the action.

    There will be the opportunity to test the algorithms on neuromorphic hardware.

    For question about this project, please contact Prof. Tirilly ([email protected]) or Dominique Chu ([email protected]).

    How to apply

    To apply please visit: https: // www. (Computer Science PhD programme).

    You will need to apply through the online application form on the main University website. Please note that you will be expected to provide personal details, education and employment history and supporting documentation (Curriculum Vitae, transcript of results, a writing sample e.g. project report, Bachelor/Master thesis, two academic references). Applications should state that you would like to be considered for this Kent-Lille studentship project and be submitted with a supporting statement from the Kent Lead Supervisor.

    From this university

    Recent blogs

    Recent news