Reproduce “Efficient Learning for Linear Properties of Bounded-Gate Quantum Circuits” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Yuxuan Du, Min-Hsiu Hsieh, Dacheng Tao
Researchers from Nanyang Technological University and Hon Hai Research Institute developed a provably efficient classical machine learning protocol for predicting linear properties of quantum circuits containing a bounded number of tunable gates. This method, which combines classical shadows with a novel trigonometric kernel, enables accurate pretraining of variational quantum algorithms and efficient characterization of circuit dynamics for up to 60 qubits.
Key insights
- While learning linear properties of bounded-gate circuits is sample-efficient (scaling linearly with the number of tunable gates `d`), it is computationally hard for classical learners, exhibiting an exponential dependency on `d`.
- The proposed kernel-based method, by utilizing classical shadows and truncated trigonometric expansions, can overcome this computational hardness and achieve polynomial runtime and memory costs under certain conditions, such as a bounded gradient norm `C`.
- Classical shadows enable the collection of quantum measurement data once, allowing subsequent flexible and efficient classical computation of expectation values for various observables without further quantum resource usage.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.