Motivated by recent progress in quantum technologies and in particular quantum software, research and industrial communities have been trying to discover new applications of quantum algorithms such as quantum optimization and machine learning. Regardless of which hardware platform these novel algorithms operate on, whether it is adiabatic or gate based, from theoretical point of view, they are performing drastically better than their classical counterparts. Two promising areas of quantum algorithms quantum machine learning and quantum optimization. These are based on performing matrix operations using quantum states and operation, in order to speed up data analysis where quantum computing can efficiently work with high dimensional vectors. Motivated by that, quantum inspired algorithms (e.g. for recommendation systems and principal component analysis) are developed to cope with high dimensionality using probabilistic techniques that are inspire from quantum computing. In this paper we review recent progress in the area of quantum inspired algorithms for low rank matrix approximation. We further explore the possibility of using parametrized complexity for such algorithms to refine practical complexity analysis. Finally, we conjecture that quantum inspired algorithms that use low rank approximation and also sample and query technique for input representations are Fixed Parameter Tractable (FPT).
翻译:受量子技术、特别是量子软件、研究和工业最近进步的启发,它们一直在试图发现量子算法的新应用,例如量子优化和机器学习。不管这些新算法是运行在哪个硬件平台上,从理论角度看,这些新算法是非对称还是基于门的,从理论角度看,它们的表现大大优于古典的对应方。两个有希望的量子算法领域量子机器学习和量子优化。这两个领域基于利用量子状态和操作进行矩阵操作,以便加速数据分析,使量子计算能够有效地与高维矢量矢量进行工作。最后,我们预测量子激励算法(例如建议系统和主要组成部分分析)是用来利用量子计算所启发的概率技术应对高维度的(例如,建议系统和主要组成部分分析)。在本文中,我们审查了量子值启发算法领域在低量子矩阵逼近法领域的最新进展。我们进一步探讨在这种算算法中使用分解复杂度复杂性来改进实际分析的可能性。最后,我们推测量子启发算算法使用低级近似和输入图的抽样和查询技术。