Trainability · Primitives

Lie-algebraic (g-sim) loss reconstruction

Theorem statement

Let $\mathfrak g=\langle\{i A_l\}\rangle_{\operatorname{Lie}}\subseteq\mathfrak u(2^n)$ be the dynamical Lie algebra generated by the skew-Hermitian generators $i A_l$ of a variational circuit (with $A_l$ Hermitian), and let $\{B_j\}_{j=1}^{\dim\mathfrak g}$ be a Hilbert–Schmidt-orthonormal basis of the Hermitian sector $i\mathfrak g$, so that $\{i B_j\}$ is a basis of $\mathfrak g$ and the quantum data $\operatorname{Tr}[\rho B_j]$ are real. For every circuit $U=\prod_k e^{i A_k}$ with Hermitian generators $A_k\in i\mathfrak g$ and every Hermitian observable $O\in i\mathfrak g$, the loss reconstructs exactly as $$\operatorname{Tr}\!\big[\,U\rho\,U^{\dagger}O\,\big]=\sum_{j=1}^{\dim\mathfrak g}\big\langle B_j,\,\mathcal H(O)\big\rangle_{\mathrm{HS}}\;\operatorname{Tr}[\rho B_j],$$ where $\mathcal H(O)=U^{\dagger}OU$ is the Heisenberg-evolved observable and $\langle B_j,\mathcal H(O)\rangle_{\mathrm{HS}}=\operatorname{Tr}[B_j\,\mathcal H(O)]$ are its basis coordinates.

Sources

  1. Lie-algebraic classical simulations for quantum computing

    Matthew L. Goh, Martin Larocca, Lukasz Cincio, M. Cerezo, Frederic Sauvage, 2023

  2. Does provable absence of barren plateaus imply classical simulability? Or, why we need to rethink variational quantum computing

    M. Cerezo, Martin Larocca, Diego Garcia-Martin, N. L. Diaz, Paolo Braccia, Enrico Fontana, Manuel S. Rudolph, Pablo Bermejo, Aroosa Ijaz, Supanut Thanasilp, Eric R. Anschuetz, Zoe Holmes, 2023

Lean context

Copy a short prompt with the import, theorem name, citations, and public source link.

Open Lean source