Implementing variational optimization on fault-tolerant quantum machines

Determine effective methods for implementing variational optimization on fault-tolerant quantum computers, addressing how such optimization should be carried out under fault-tolerant resource constraints.

Background

The paper argues that variational approaches remain relevant beyond NISQ, for example to prepare good initial states for non-variational subroutines like phase estimation. However, fault-tolerant execution introduces new constraints, such as fine-grained rotation synthesis potentially incurring high T-gate counts and slower logical clock rates that make shot economy crucial.

While techniques like error mitigation for compilation and algorithmic errors, amplitude amplification, and classical shadows may help, concrete best practices for fault-tolerant variational optimization remain unsettled and require further investigation.

References

Nonetheless, there remain open questions on how to best implement variational optimization in fault tolerant machines but the field is gaining traction.

— Myths around quantum computation before full fault tolerance: What no-go theorems rule out and what they don't  (2501.05694 - Zimborás et al., 10 Jan 2025) in Subsection 'Variational quantum algorithms', Subsubsection 'Variational quantum algorithms beyond NISQ'

Nevertheless, a significant disadvantage lies in the need to determine beforehand the appropriate parameters of the quantum circuit that maps the inputs to quantum states, and it is not yet clear whether these parameters can be effectively obtained using a variational technique.

— Lazy training of quantum physics informed neural networks  (2609.19239 - Hernandez et al., 16 Sep 2026) in Section 1, Introduction