Projet de recherche doctoral numero :8334


Date depot: 15 avril 2022
Titre: Robust and Scalable Noise Mitigation for Quantum Computing
Directrice de thèse: Elham KASHEFI (LIP6)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Modèles de calcul, preuve, vérification

Resumé: Background and Previous Research Quantum computing promises to transform society for the better, with problem-solving capability fundamentally out of reach without it, and new possibilities for secure networks. To fulfill this potential, it needs to be scaled up by orders of magnitude. Quantum software has the potential to do this. It is the fastest-growing segment within all of quantum technology. Correcting errors due to noise in quantum circuits run on current and near-term quantum hardware is essential for any convincing demonstration of quantum advantage. Indeed, in many cases it has been shown that noise renders quantum circuits efficiently classically simulable, thereby destroying any quantum advantage potentially offered by an ideal (noiseless) implementation of these circuits. There exists a wide range of error correction and error mitigation techniques to suppress the noise affecting the implementation of a given quantum program on target quantum hardware. Hardware agnostic solutions suffer huge overhead making them irrelevant for near- and medium- term quantum devices. On the other hand, various tailor-made solutions address specific noise models based on ad hoc assumptions making them unsuitable for delivering a resilient quantum platform. This project aims to address the challenge of delivering quantum computing that is correct, resilient and trustworthy, by bringing together cutting edge quantum information and quantum cryptography with classical traditional methods that have proven effective in verification. Research Vision The overall research vision is based on the improvement of the performance of quantum error mitigation (QEM) techniques in terms of their scalability and noise assumptions. This is also aimed to make the established quantum-classical hybrid algorithms and quantum verification protocols tolerant to noise. We aim to generalize and improve the QEMs themselves, to combine the QEMs with quantum algorithms and protocols in a better way, and to evaluate the actual performance on the available quantum devices. The guiding principle is to develop techniques capable of correcting a significant amount of noise in the outputs of quantum computations such that their implementation can be performed on current and near-term quantum hardware. The first step in error mitigation is a good understanding of the imperfections of the hardware, i.e. to obtain a detailed noise model that incorporates and characterizes correlated errors and cross-talk errors, going beyond the simplified single- and two-qubit models. As this characterisation is intractable analytically as the hardware scales up, we will adapt techniques from (classical) machine learning. Once a good noise model is in place, the second task is to suppress errors. We will use two methods: (1) increase the number of samples (runs) of a noisy quantum circuit, instead of encoding qubits of this circuit as logical qubits; (2) Correct the output statistics of a quantum circuit, rather than actively correcting errors affecting gates, preparations, and measurements. An accurate noise model facilitates a bespoke approach to error suppression in both cases. Classical hardware verification frameworks only certify a correct outcome in a passive way. In a quantum setting we can also separate noisy samples from the mix. We will compose various error mitigation approaches with quantum certification techniques using classical machine learning or statistical analysis.

Doctorant.e: Yang Bo