Synergistic pretraining of parametrized quantum circuits via tensor networks

Area of Science:

• Quantum Computing
• Computational Chemistry
• Materials Science
• Machine Learning

Background:

• Parametrized quantum circuits (PQCs) are a key approach for leveraging current quantum hardware.
• Barren plateaus and the challenge of outperforming classical algorithms hinder PQC development.
• Efficiently optimizing PQCs is crucial for realizing quantum advantage.

Purpose of the Study:

• To introduce a synergistic framework combining classical and quantum resources for PQC optimization.
• To address and mitigate the issues of barren plateaus and classical performance limitations in PQCs.
• To demonstrate a pathway for achieving practical quantum advantage using hybrid methods.

Main Methods:

• A synergistic framework is proposed, integrating classical tensor network computations with PQCs.
• Classical resources are used to generate an initial high-quality solution via tensor networks.
• This classical output is then converted into a PQC for further refinement using quantum computation.

Main Results:

• The synergistic approach effectively mitigates barren plateaus in PQCs for systems up to 100 qubits.
• Performance improves with the increased application of either classical or quantum resources.
• Numerical evidence supports the framework's ability to overcome optimization challenges.

Conclusions:

• Classical simulation methods can be integral to achieving quantum advantage, not merely obstacles.
• The synergistic approach offers a promising direction for developing practically useful quantum algorithms.
• Hybrid classical-quantum strategies are essential for advancing the field of quantum computation.