Connectionists: Two PhD positions - MSCA Doctoral Network “ELEVATE” (101227453)

[Drion Guillaume]

The Faculty of Applied Sciences at the University of Liege has two open PhD positions connected to the MSCA Doctoral Network “ELEVATE” (101227453) (R15 and R16 in https://www.elevate-dn.eu/). The proposed PhD projects will be hosted by the ULiège Neuroengineering Lab, whose mission is to understand biological intelligence and leverage this understanding to design more resilient and green machines and computing architectures.

Deadline: 05/03/2026. Detailed information can be found in the links below.

Project R15: Leveraging neuron rhythmic transitions in SNNs to develop novel bio-inspired learning rules
<https://www.neuroengineering.uliege.be/upload/docs/application/pdf/2026-02/job_vacancy_msca_docnet_elevate_r15.pdf>
[preview.png]
job_vacancy_msca_docnet_elevate_r15<https://www.neuroengineering.uliege.be/upload/docs/application/pdf/2026-02/job_vacancy_msca_docnet_elevate_r15.pdf>
PDF Document · 295 KB<https://www.neuroengineering.uliege.be/upload/docs/application/pdf/2026-02/job_vacancy_msca_docnet_elevate_r15.pdf>

One major challenge facing neuromorphic chips is the ability to improve their performance and learn novel tasks online while interacting with the environment. In biological brains, such online learning is performed through the use of synaptic plasticity rules that depend on spike timing between presynaptic and postsynaptic neurons. Although mathematical models of such rules have been available for a while, they have been shown to be difficult to parametrize and they underperform in SNNs. Furthermore, learning and behavior happening in parallel, both often interfere with each other. In this project, we aim to develop novel bio -inspired plasticity rules to allow for online learning in neuromodulable SNNs. We will combine rhythmic transitions observed in biological systems with novel local learning rules to achieve robust online memory consolidation in neuromorphic systems.


Project R16: Neuromodulation for stable learning in recurrent neural networks
<https://www.neuroengineering.uliege.be/upload/docs/application/pdf/2026-02/job_vacancy_msca_docnet_elevate_r16.pdf>
[preview.png]
job_vacancy_msca_docnet_elevate_r16<https://www.neuroengineering.uliege.be/upload/docs/application/pdf/2026-02/job_vacancy_msca_docnet_elevate_r16.pdf>
PDF Document · 294 KB<https://www.neuroengineering.uliege.be/upload/docs/application/pdf/2026-02/job_vacancy_msca_docnet_elevate_r16.pdf>

Recurrent neural networks (RNNs) are emerging as a powerful model for efficient processing of temporal data in low-footprint edge devices. However, while local learning rules have recently been shown to enable high-performance learning with RNNs, they are notoriously unstable. To solve this problem, this project aims to develop stable local learning rules for RNNs by taking inspiration from a central aspect of stable learning in the brain, rhythmic activity, thereby outlining a flexible spectrum from regulation to adaptation.

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20260205/491be28d/attachment-0001.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: preview.png
Type: image/png
Size: 340740 bytes
Desc: preview.png
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20260205/491be28d/attachment-0002.png>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: preview.png
Type: image/png
Size: 313036 bytes
Desc: preview.png
URL: <http://mailman.srv.cs.cmu.edu/pipermail/connectionists/attachments/20260205/491be28d/attachment-0003.png>