.png){kind=icon}
返回
Using quantum transport networks for classification: A path toward quantum computing for machine learning
.png){kind=icon}
PRE
AI
DOI：10.1103/PhysRevApplied.22.014041.png)
.png){kind=icon}
原文求助.png){kind=icon}
分享.png){kind=icon}
收藏
摘要
En 中文
Classification, the computational process of categorizing an input into preexisting classes, is now a cornerstone in modern computation in the era of machine learning. Here, we propose an approach for a quantum physical computer; a quantum classifier, based on quantum transport of particles in a trained quantum network. The classifier is based on sending a quantum particle into a network and measuring the exit point of the particle, which serves as a class and can be determined by changing the network parameters, differing from standard quantum computers as no gate operations are required to perform the computation. Using this scheme, we demonstrate three examples of classification. In the first, wave functions are classified according to their overlap with predetermined (random) groups. In the second, we classify wave functions according to their level of localization. Both examples use small training sets and achieve over 95% precision and recall. The third classification scheme is a real-world problem, concerning classification of catalytic aromatic aldehyde substrates according to their reactivity. Using experimental data, the quantum classifier reaches an average 86% classification accuracy. We show that the quantum classifier outperforms its classical counterpart for these examples and demonstrates clear advantage, especially in the regime of small data. These results pave the way for a classification scheme that can be implemented as an algorithm and potentially realized experimentally on quantum hardware.
Keyword:
INVERSE PARTICIPATION RATIO
FLUCTUATIONS
期刊
IF:
4.4
论文数:
7.1K
被引数:
2.8W
机构
引用论文
分享收藏Enhancing coherent transport in a photonic network using controllable decoherence使用可控退相干增强光子网络中的相干传输.png)
NATURE COMMUNICATIONS
OAAI
NATURE COMMUNICATIONSIF15.7
OAAI分享收藏Molecular rectifier composed of DNA with high rectification ratio enabled by intercalationNATURE CHEMISTRY
NATURE CHEMISTRYIF20.2
PREAI
分享收藏Physics-based neural network for non-invasive control of coherent light in scattering mediaOPTICS EXPRESS
OAAI
OPTICS EXPRESSIF3.3
OAAI分享收藏
查看更多内容