arXivDaily arXiv每日学术速递 周一至周五更新
arXiv 2405.01780quant-phcs.LG
收藏

量子机器学习:量子核方法

Quantum Machine Learning: Quantum Kernel Methods

QSPectral Systems

Sanjeev Naguleswaran

AI总结:

本文研究量子核方法在支持向量机和卷积神经网络中的应用,通过数据依赖的投影量子核展示量子优势,并探索其作为CNN特征提取层的扩展。

摘要:

基于量子核方法的量子算法此前已被研究[1]。量子优势源于这样一个事实:可以构造一族数据集,对于这些数据集,只有量子处理能够识别其内在的标注模式,而对于经典计算机而言,该数据集看起来像噪声。这是因为该算法利用了循环群中对数计算的内在效率。离散对数问题是量子计算相对于经典计算的一个众所周知的优势:其中可以用单一数学运算生成群的所有成员。核方法是经典机器学习中一种强大且流行的技术。使用只能在量子计算机上高效计算的量子特征空间,有可能带来量子优势。在本文中,我们首先旨在描述将这种核方法应用于经典支持向量机(SVM)算法的量子版本,以确定实现量子优势的条件。研究表明,一种数据依赖的投影量子核相比经典核具有显著优势。此外,我们展示了关于将量子核作为特征提取层扩展到卷积神经网络(CNN)中的研究结果和构想,CNN是深度学习应用中被广泛使用的架构。

英文摘要:

Quantum algorithms based on quantum kernel methods have been investigated previously [1]. A quantum advantage is derived from the fact that it is possible to construct a family of datasets for which, only quantum processing can recognise the intrinsic labelling patterns, while for classical computers the dataset looks like noise. This is due to the algorithm leveraging inherent efficiencies in the computation of logarithms in a cyclic group. The discrete log problem.is a well-known advantage of quantum vs classical computation: where it is possible to generate all the members of the group using a single mathematical operation. Kernel methods are a powerful and popular technique in classical Machine Learning. The use of a quantum feature space that can only be calculated efficiently on a quantum computer potentially allows for deriving a quantum advantage. In this paper, we intend to first describe the application of such a kernel method to a Quantum version of the classical Support Vector Machine (SVM) algorithm to identify conditions under which, a quantum advantage is realised. A data dependent projected quantum kernel was shown to provide significant advantage over classical kernels. Further, we present results of investigations and ideas pertaining to extending the use of quantum kernels as a feature extraction layer in a Convolutional Neural Networks (CNN) that is a widely used architecture in deep-learning applications.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑