Reproduce “Hierarchical Learning for Quantum ML: Novel Training Technique for Large-Scale Variational 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
Hrant Gharibyan, Vincent Su, Hayk Tepanyan
BlueQubit researchers developed a hierarchical learning technique enabling the efficient training of variational quantum circuits (VQCs) up to 27 qubits, practically circumventing barren plateaus for Quantum Circuit Born Machines. This method, combined with an optimized adjoint derivative, achieved approximately 4% Total Variation distance for a 3D multivariate Gaussian and demonstrated improved accuracy on IBM quantum hardware with error mitigation.
Key insights
- Hierarchical learning provides a practical strategy to mitigate barren plateaus in VQC training, enabling successful optimization of circuits with up to 27 qubits.
- Circuit connectivity, specifically 2D grid topology, significantly impacts a VQC's expressibility and trainability for distribution loading tasks, outperforming 1D line connectivity.
- The extended adjoint derivative method is crucial for scaling VQC training on classical simulators, offering substantial speed-ups (e.g., quadratic in parameters) for complex loss functions.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.