Reproduce “Quantum Natural Gradient” 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
James Stokes, Josh Izaac, Nathan Killoran, Giuseppe Carleo
The paper introduces the Quantum Natural Gradient (QNG), an optimization algorithm that leverages the Fubini-Study metric to guide variational quantum algorithms (VQAs) along the information geometry of quantum state space. Numerical experiments demonstrate that QNG consistently achieves faster convergence in fewer iterations compared to standard gradient descent and Adam optimizers for VQAs, particularly as qubit count and circuit depth increase.
Key insights
- The Fubini-Study metric tensor, derived from the Quantum Geometric Tensor (QGT), provides the natural Riemannian metric for the complex projective space of pure quantum states, extending the concept of information geometry to the quantum domain.
- The Quantum Natural Gradient (QNG) optimizer, when applied to VQAs, achieves optimization by effectively performing imaginary-time evolution within the variational subspace, ensuring parametrization invariance.
- Practical approximations of the QGT (block-diagonal and diagonal) can be efficiently computed on quantum hardware, allowing the QNG to be implemented with a reasonable quantum resource overhead.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.