Approximate Quantum State Preparation Through Proximal Policy Optimization

2026-07-23 · Source: Machine Learning · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation · Depth: Expert, quick

Summary

A quantum architecture search framework is proposed for approximate quantum state preparation (QSP), a challenging task due to the exponentially growing search space with qubit count. This framework employs deep reinforcement learning, specifically an agent based on proximal policy optimization, to identify optimal quantum circuits. The agent's objective is to achieve the best possible approximation of a target state while simultaneously minimizing the number of gates. The process involves the agent iteratively appending new gates and recomputing fidelity. Experiments conducted with 2 to 5 qubits, using predefined states like Bell, GHZ, W, and Dicke states, alongside completely random states, demonstrated the framework's capability to achieve approximation errors as low as 10e-14.

Key takeaway

For research scientists focused on quantum computing and circuit optimization, this work demonstrates a powerful approach to quantum state preparation. If you are designing quantum circuits for specific or random states, consider integrating deep reinforcement learning with Proximal Policy Optimization. This method efficiently navigates the complex, exponentially growing search space, enabling the discovery of circuits that achieve high approximation fidelity (e.g., 10e-14) while minimizing gate usage, thereby improving circuit efficiency.

Key insights

Deep reinforcement learning via Proximal Policy Optimization effectively addresses the exponential search space in quantum state preparation.

Principles

• Quantum state preparation faces an exponentially growing search space.
• Optimizing quantum circuits requires balancing state fidelity with gate count minimization.

Method

An agent iteratively appends quantum gates to a circuit, recomputes fidelity, and uses Proximal Policy Optimization to learn optimal gate sequences for target state approximation.

In practice

• This framework can prepare specific quantum states (e.g., Bell, GHZ, W, Dicke).
• It can approximate arbitrary random states with high precision on 2-5 qubit systems.

Topics

• Quantum State Preparation
• Proximal Policy Optimization
• Quantum Architecture Search
• Deep Reinforcement Learning
• Quantum Computing
• Quantum Circuits

Best for: AI Scientist, Research Scientist

Related on AIssential

• Designing better quantum circuits with AI — AI-driven reinforcement learning can significantly optimize quantum circuit design by reducing gate count.
• Graph Reinforcement Learning for Calibration-Aware Quantum Circuit Routing — Integrating real-time hardware calibration into quantum circuit routing via graph reinforcement learning significantly enhances circuit fidelity.
• Enhanced Reinforcement Learning-based Process Synthesis via Quantum Computing — Quantum reinforcement learning with state encoding enhances scalability for process synthesis problems.
• One More Time: Revisiting Neural Quantum States from a Reinforcement Learning Perspective — Variational energy minimization for NQS can be optimized via a trust-region policy-gradient approach, improving stability and scale.
• Cautious optimism for deep parameterized quantum circuits — Gradient-based parameterized quantum circuits can exhibit double descent, improving generalization with increased model size, challenging traditional views.
• AlphaCNOT: Learning CNOT Minimization with Model-Based Planning — AlphaCNOT combines RL with Monte Carlo Tree Search for efficient CNOT gate minimization in quantum circuits.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Machine Learning.