PhD thesis: Training of Quantum Neural Networks via Multiplexed Perturbation (M/F)

CNRS; France; Sur place; EUR 60 000 - 80 000; Plein temps

Résumé du poste

A leading research institution in France is seeking a PhD candidate for a project focused on quantum neural networks. The successful applicant will join the Neuromorphic Computing team, collaborate closely with peers, and implement innovative approaches for data learning in quantum systems. A strong background in physics and experience in programming are essential for this role. The position offers a full-time contract starting from January 2026.

Qualifications

• Strong background in quantum physics and neural networks is essential.
• Experience in experimental physics is preferred.

Responsabilités

• Adapt a novel gradient estimation method to bosonic networks.
• Design and implement a coupled circuit for experiments.
• Demonstrate learning of coupling parameters experimentally.

Connaissances

Knowledge of quantum neural networks

Experience with quantum computing techniques

Proficiency in programming

Formation

Master's degree in Physics or related field

Description du poste

Organisation/Company CNRS Department Laboratoire Albert Fert Research Field Physics » Condensed matter properties Physics » Solid state physics Physics » Surface physics Researcher Profile First Stage Researcher (R1) Country France Application Deadline 17 Dec 2025 - 23:59 (UTC) Type of Contract Temporary Job Status Full-time Hours Per Week 35 Offer Starting Date 1 Jan 2026 Is the job funded through the EU Research Framework Programme? Horizon Europe - ERC Is the Job related to staff position within a Research Infrastructure? No

Offer Description

This PhD project is part of the ERC project QDYNNET – Quantum Dynamical Neural Networks, led by Danijela Marković.

The successful candidate will join the Neuromorphic Computing team at the Albert Fert Laboratory (CNRS, Thales, Université Paris-Saclay), and will work in close collaboration with two other PhD students recruited within the project.

The thesis will be supervised by Danijela Marković and Julie Grollier (CNRS).

Quantum neural networks (QNNs) are attracting increasing attention for their ability to project data into a high-dimensional Hilbert space, where they can become more easily separable. They also offer the possibility of learning directly from quantum data, thanks to their natural compatibility with other quantum systems that may generate such data.

The most common approach relies on variational quantum circuits based on qubits. However, training such networks poses major challenges: the efficient estimation of gradients of the outputs with respect to internal parameters, and the problem of barren plateaus (vanishing gradients), due to information dilution in large Hilbert space.

In our group, we are exploring an alternative approach based on coupled bosonic modes [1,2], where information is encoded in coherent states and manipulated through continuous operations such as displacement, squeezing, and parametric coupling. This method preserves structure within the Hilbert space, which may mitigate information dilution and facilitate learning.

Recently, a novel multiplexed perturbation method has been proposed, allowing simultaneous estimation of gradients with respect to multiple parameters by modulating them sinusoidally at distinct frequencies [3]. In quantum systems, by choosing the appropriate perturbation amplitude, the gradient can be obtained exactly, without approximation [4].

The goal of this PhD project is to:

• adapt this method to parametric bosonic networks,
• design and experimentally implement a tunable four-mode coupled circuit,
• and demonstrate experimental learning of coupling parameters using this approach.

References:

• 1. Dudas, J. et al. Quantum reservoir computing implementation on coherently coupled quantum oscillators. Npj Quantum Inf. 9, 64 (2023).
• 2. Dudas, J., Carles, B., Gouzien, E., Grollier, J. & Marković, D. Training the parametric interactions in an analog bosonic quantum neural network with Fock basis measurement. Preprint at https://doi.org/10.48550/arXiv.2411.19112 (2024).
• 3. McCaughan, A. N. et al. Multiplexed gradient descent: Fast online training of modern datasets on hardware neural networks without backpropagation. APL Mach. Learn. 1, 026118 (2023).
• 4. Hoch, F. et al. Variational approach to photonic quantum circuits via the parameter shift rule. Phys. Rev. Res. 7, 023227 (2025).