[Paper Review] Predicting Good Quantum Circuit Compilation Options

Nils Quetschlich, Lukas Burgholzer+1 more|arXiv (Cornell University)|Oct 14, 2022

Quantum Computing Algorithms and Architecture43 references4 citations

TL;DR

This paper proposes a supervised machine learning framework that predicts the optimal combination of qubit technology, device, compiler, and settings for compiling quantum circuits—demonstrating it achieves the best compilation option for over 75% of unseen circuits and reduces median compilation time by more than an order of magnitude. The approach treats quantum circuit compilation selection as a classification task, leveraging trained models to automate expert-level decision-making for non-expert users.

ABSTRACT

Any potential application of quantum computing, once encoded as a quantum circuit, needs to be compiled in order to be executed on a quantum computer. Deciding which qubit technology, which device, which compiler, and which corresponding settings are best for the considered problem -- according to a measure of goodness -- requires expert knowledge and is overwhelming for end-users from different domains trying to use quantum computing to their advantage. In this work, we treat the problem as a statistical classification task and explore the utilization of supervised machine learning techniques to optimize the compilation of quantum circuits. Based on that, we propose a framework that, given a quantum circuit, predicts the best combination of these options and, therefore, automatically makes these decisions for end-users. Experimental evaluations show that, considering a prototypical setting with 3000 quantum circuits, the proposed framework yields promising results: for more than three quarters of all unseen test circuits, the best combination of compilation options is determined. Moreover, for more than 95% of the circuits, a combination of compilation options within the top-three is determined -- while the median compilation time is reduced by more than one order of magnitude. Furthermore, the resulting methodology not only provides end-users with a prediction of the best compilation options, but also provides means to extract explicit knowledge from the machine learning technique. This knowledge helps in two ways: it lays the foundation for further applications of machine learning in this domain and, also, allows one to quickly verify whether a machine learning algorithm is reasonably trained. The corresponding framework and the pre-trained classifier are publicly available on GitHub (https://github.com/cda-tum/MQTPredictor) as part of the Munich Quantum Toolkit (MQT).

Motivation & Objective

• To address the overwhelming complexity faced by non-expert users in selecting optimal quantum compilation options across diverse qubit technologies, devices, compilers, and settings.
• To automate the decision-making process for quantum circuit compilation by transforming it into a supervised machine learning classification task.
• To reduce the time and expertise required for selecting high-performing compilation configurations, especially for end-users without quantum computing backgrounds.
• To provide interpretable insights from the trained model, enabling verification of model behavior and guiding future ML applications in quantum compilation.
• To create a scalable, extensible framework that can adapt to new devices, compilers, and calibration data without full retraining from scratch.

Proposed method

• The framework treats the selection of compilation options as a multi-class classification problem, where the input is a quantum circuit and the output is the best combination of qubit technology, device, compiler, and settings.
• A dataset of 3,000 quantum circuits from the MQT Bench library is compiled using 30 different combinations of options (2 technologies, 5 devices, 2 compilers, 6 settings), generating labeled training data.
• A supervised machine learning classifier is trained on this data to predict the optimal compilation configuration for new, unseen circuits.
• The model is designed to be interpretable, allowing extraction of explicit knowledge from feature importance and decision paths to validate training quality.
• The framework supports incremental updates: re-evaluation of existing compiled circuits with new calibration data or new compilation options enables retraining without recompilation.
• A warm-start retraining strategy is employed, using previously trained model parameters as initialization to improve efficiency when updating with new data.

Experimental results

Research questions

• RQ1 Can a machine learning model accurately predict the best compilation configuration (qubit technology, device, compiler, settings) for a given quantum circuit?
• RQ2 How does the model’s performance scale across unseen quantum circuits in terms of top-1 and top-3 prediction accuracy?
• RQ3 To what extent can the model reduce compilation time compared to exhaustive manual search over all configuration combinations?
• RQ4 Can the model provide interpretable insights into its predictions, enabling validation and guiding future research?
• RQ5 How efficiently can the model be updated to incorporate new devices, compilers, or calibration data without full retraining?

Key findings

• The proposed framework predicts the best compilation option (top-1) for more than 75% of unseen test circuits, demonstrating strong generalization.
• For over 95% of the circuits, the model identifies a compilation configuration within the top-three best options, indicating high reliability.
• The median compilation time is reduced by more than one order of magnitude compared to exhaustively testing all 30 configuration combinations and selecting the best result.
• The framework supports incremental updates: existing compiled circuits are persistently stored, enabling re-evaluation with new calibration data without recompilation.
• The model’s interpretability allows users to extract explicit knowledge from predictions, enabling validation of model behavior and guiding future ML applications in quantum compilation.
• The open-source framework and pre-trained classifier are publicly available on GitHub as part of the Munich Quantum Toolkit (MQT), facilitating adoption and extension.