Exploratory action

MARCQ

Hybrid methods combining Reinforcement Learning and PDE Optimal Control methods for Quantum Computing
Hybrid methods combining Reinforcement Learning and PDE Optimal Control methods for Quantum Computing

This project pertains to quantum computing: we are interested in the possibility of encoding a logic gate, such as the Hadamard gate or the "not" gate, using qudit systems with the aid of controls. This represents a promising alternative to the usual approaches that use qubit systems. These issues are addressed using optimal control problems. The underlying dynamic model is given by the Lindblad equation. This question is challenging due to the emergence of a physical phenomenon called decoherence, which opposes the action of the control. We aim to study the influence of parameters ensuring the effectiveness of the controls, the dependence on the system dimension, and to develop a numerical study based on the combination of traditional fixed-point algorithms and learning methods adapted to the problem and its potentially large dimension depending on the molecules studied. The ultimate goal is the experimental implementation of the obtained strategies, in collaboration with IPCMS.

Inria teams involved
TONUS

Contacts

Yannick Privat

Scientific leader