Speaker
Description
Efficient measurement of quantum states is a fundamental task in quantum computing. In particular, for many near-term quantum algorithms, a key challenge is how to accurately estimate a large number of non-commuting Pauli observables with as few measurements as possible. One common strategy is to group observables into commuting sets, but measuring general commuting groups often requires deep circuits that are challenging to implement on near-term devices and remain costly in early fault-tolerant regimes. A more hardware-friendly alternative is to restrict to single-qubit rotations and qubit-wise commuting groups, but this often leads to smaller measurement groups and higher sampling cost.
In this talk, I will present a new measurement scheme based on generative learning. Rather than constructing explicit measurement groupings by solving NP-hard combinatorial optimization problems or performing nontrivial derandomization of classical shadows, our approach directly learns an ensemble of measurement circuits optimized for a given set of observables and resource constraints, such as the number of measurements and allowed circuit depth. Numerical results demonstrate systematic performance gains over state-of-the-art methods as well as promising generalization beyond the training regime. Our learning-based approach provides a flexible and general framework for practical measurement design in both near-term and early fault-tolerant settings, where quantum resources remain limited and imperfect.