SQuASH dataset
收藏资源简介:
SQuASH数据集是一个为量子架构搜索(QAS)方法设计的大型数据集,包含超过300万个独特的参数化量子电路(PQC)。该数据集由Fraunhofer FOKUS和柏林自由大学的研究团队创建,旨在通过使用代理模型来加速QAS方法的评估,并促进QAS方法的可重复性和进一步研究。数据集涵盖了两个不同的任务:量子态制备和数据点分类,每个任务都有预定义的搜索空间和预训练的代理模型,以预测优化后的PQC性能。数据集的创建过程涉及使用强化学习、遗传算法和随机搜索策略来探索预定义的搜索空间,并生成和优化候选PQC。SQuASH数据集适用于评估各种QAS方法,特别是那些将参数优化作为搜索过程内循环的方法。
The SQuASH dataset is a large-scale dataset designed for quantum architecture search (QAS) methods, containing over 3 million unique parameterized quantum circuits (PQCs). Developed by a research team from Fraunhofer FOKUS and the Free University of Berlin, this dataset aims to accelerate the evaluation of QAS approaches via surrogate models, while facilitating the reproducibility and further research of QAS methods. The dataset covers two distinct tasks: quantum state preparation and data point classification. Each task is equipped with a predefined search space and a pre-trained surrogate model to predict the performance of optimized PQCs. The creation of the SQuASH dataset involved using reinforcement learning, genetic algorithms, and random search strategies to explore the predefined search space, as well as generate and optimize candidate PQCs. The SQuASH dataset is suitable for evaluating various QAS methods, particularly those that treat parameter optimization as an inner loop of the search process.

- 1Benchmarking Quantum Architecture Search with Surrogate AssistanceFraunhofer FOKUS, Berlin, Germany; Freie Universit¨at Berlin, Berlin, Germany · 2025年



