Analysis of Precision Vectors for Ising-Based Linear Regression
Analysis of Precision Vectors for Ising-Based Linear Regression
复制标题
基于 Ising 的线性回归的精度向量分析
DOI：
10.1007/978-3-031-29927-8_20
复制
发表时间：
2023
期刊：
影响因子：
--
通讯作者：
Kobayashi Hiroaki
中科院分区：
文献类型：
--
作者：
Aoyama Kaho;Komatsu Kazuhiko;Kumagai Masahito;Kobayashi Hiroaki
量子计算作为一种新的计算原理受到了广泛的关注。特别是，利用统计力学的伊辛模型的退火机器是新兴的和可行的下一代计算技术。退火炉可以解决经典计算原理中难以解决的组合优化问题。目前，基于ising的算法正在大力发展，通过将其作为组合优化问题解决来执行各种应用，例如机器学习。另一方面，机器学习应用中使用的传感器数据和模拟数据等数据量正在急剧增加。处理大量数据变得很困难。特别是，由于退火机器的量子比特数量有限，因此强烈需要独立于数据大小的基于ising的算法。本文主要研究基于ising的线性回归方法，该方法利用精度向量代替目标数据的每个数据元素。虽然精确向量的使用是减少量子位数量与数据量的关键，但诸如精确向量的多少个元素是必要的以及这些元素如何设置等细节尚未明确。本文通过实证方法对基于ising的线性回归进行性能评价，讨论了精度向量的特征。实验结果表明，考虑输入数据集，选择合适的精度向量可以提高线性回归的质量。
Quantum computing has been much attention as one of the new computational principles. In particular, annealing machines that use the Ising model of statistical mechanics are emerging and feasible next-generation computational technology. Annealing machines can solve combinatorial optimization problems that have been considered difficult to solve in classical computing principles. Currently, Ising-based algorithms are being vigorously developed to perform various applications such as machine learning by solving them as combinatorial optimization problems. On the other hand, the amount of data, such as sensor data and simulation data, used in machine learning applications is drastically increasing. It becomes difficult to handle large amounts of data. In particular, since the number of qubits of annealing machines is limited, Ising-based algorithms independent of data size are strongly required.This paper focuses on Ising-based linear regression that utilizes a precision vector instead of each data element of target data. Although the use of a precision vector is a key point that can reduce the number of qubits against the amount of data, detail such as how many elements of a precision vector is necessary and how these elements are set to be is not clarified yet. This paper discusses the characteristics of a precision vector through performance evaluation of Ising-based linear regression with a practical data set in empirical ways.The experimental results show that a proper precision vector considering the input data set can improve the quality of the linear regression.