Synthesis of discrete-continuous quantum circuits with multimodal diffusion models

2025-06-02

AI Summary

Quantum circuit compilation via the search-and-optimization paradigm suffers from prohibitive hardware overhead and poor scalability. Method: This paper introduces the first multimodal denoising diffusion model capable of jointly generating discrete gate structures and continuous gate parameters. It proposes a novel dual-path independent diffusion mechanism that enables end-to-end differentiable modeling of symbolic gate-sequence encodings and parameterized gate embeddings—eliminating reliance on quantum hardware execution or costly classical simulation. Contribution/Results: The method achieves low compilation error on 2–6-qubit benchmarks, with generation speed accelerated by three orders of magnitude over conventional approaches. We also construct the first large-scale quantum circuit dataset and derive novel synthetic heuristic rules from it.

Technology Category

• Machine Learning: Quantum Machine Learning
• Search and Optimization: Mixed Discrete/Continuous Search
• Computer Vision: Diffusion Models for Vision

Application Category

• Security and Privacy: Large-scale security measurements
• Economics, Online Markets and Human Computation: LLM based quality controls for crowd work
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Efficiently compiling quantum operations remains a major bottleneck in scaling quantum computing. Today's state-of-the-art methods achieve low compilation error by combining search algorithms with gradient-based parameter optimization, but they incur long runtimes and require multiple calls to quantum hardware or expensive classical simulations, making their scaling prohibitive. Recently, machine-learning models have emerged as an alternative, though they are currently restricted to discrete gate sets. Here, we introduce a multimodal denoising diffusion model that simultaneously generates a circuit's structure and its continuous parameters for compiling a target unitary. It leverages two independent diffusion processes, one for discrete gate selection and one for parameter prediction. We benchmark the model over different experiments, analyzing the method's accuracy across varying qubit counts, circuit depths, and proportions of parameterized gates. Finally, by exploiting its rapid circuit generation, we create large datasets of circuits for particular operations and use these to extract valuable heuristics that can help us discover new insights into quantum circuit synthesis.

Problem

• Efficiently compiling quantum operations remains a bottleneck
• Current methods have long runtimes and high costs
• Machine-learning models are restricted to discrete gate sets

Innovation

• Multimodal diffusion model for quantum circuits
• Combines discrete and continuous diffusion processes
• Generates circuit structure and parameters simultaneously