基于张量网络与纠缠理论的高效可解释量子机器学习算法

批准号:
12004266
项目类别:
青年科学基金项目
资助金额:
24.0 万元
负责人:
冉仕举
依托单位:
学科分类:
强关联体系
结题年份:
2023
批准年份:
2020
项目状态:
已结题
项目参与者:
冉仕举
国基评审专家1V1指导 中标率高出同行96.8%
结合最新热点,提供专业选题建议
深度指导申报书撰写,确保创新可行
指导项目中标800+,快速提高中标率
微信扫码咨询
中文摘要
本项目探索基于张量网络与纠缠理论的高效可解释量子机器学习算法。作为量子物理与机器学习的交叉领域,量子机器学习算法最近几年取得显著进展,在研究强关联格点模型及其量子相的高效识别等研究方向具有重要应用,为量子新奇物态的探索提供了全新思路。但是由于传统机器学习方法通常具备极高的复杂度,导致其模型原理及计算结果往往缺失可解释性,这阻碍了其在解决具有挑战性的量子物理问题上的应用。因此,发展新的可解释机器学习模型成为一个关键性的科学问题。针对传统机器学习模型的缺陷,本项目将基于张量网络与量子纠缠理论,发展新的高效可解释机器学习模型及算法,实现强关联自旋系统的量子相识别,包括磁有序相等可由序参量描述的相,以及自旋液体等不能由序参量描述的非传统相。本项目的研究将推动基于量子理论的人工智能方法的理解与应用。
英文摘要
In this project, we will explore efficient and interpretable quantum machine learning (ML) algorithms based on tensor networks (TN's) and entanglement theories. As an interdisciplinary field of quantum physics and machine learning, astonishing progresses on quantum ML have been made in recent years. It has important applications in studying strongly-correlated lattice models and the recognitions of their quantum phases, providing new paths to investigate novel states of matter. However, due to the extremely high complexity, the conventional ML models usually suffer from the lack of interpretability to either their basic principles or results. The non-interpretability hinders the applications of such models to solving the challenging problems in quantum physics. Therefore, developing novel interpretable ML models is of great importance. Regarding the drawbacks of the conventional ML models, this project aims at developing interpretable machine learning models and algorithms based on tensor network and entanglement theories, and providing efficient solutions for recognizing the quantum phases of strongly-correlated spin models. The recognitions will be implemented on the conventional phases that can be described by order parameters such as magnetically-ordered phases, and the unconventional phases that cannot be described by any local order parameters such as spin liquids. This project is expected to advance the understanding and application of the ML methods that are based on quantum theories.
期刊论文列表
专著列表
科研奖励列表
会议论文列表
专利列表
DOI:--
发表时间:2023
期刊:Quantum Science and Technology
影响因子:6.7
作者:Ding-Zu Wang;Guo-Feng Zhang;Maciej Lewenstein;Shi-Ju Ran
通讯作者:Shi-Ju Ran
DOI:10.1103/PhysRevE.103.013313
发表时间:2021
期刊:Physical Review E
影响因子:--
作者:Yuan Yang;Zhengchuan Wang;Shi-Ju Ran;Gang Su
通讯作者:Gang Su
DOI:10.1088/0256-307x/38/11/110301
发表时间:2020-12
期刊:Chinese Physics Letters
影响因子:3.5
作者:Xinran Ma;Z. C. Tu;Shi-Ju Ran
通讯作者:Xinran Ma;Z. C. Tu;Shi-Ju Ran
DOI:10.1103/physrevb.105.165116
发表时间:2022-01
期刊:Physical Review B
影响因子:3.7
作者:Rui Hong;Yanli Xiao;Jie Hu;A. Ji;Shi-Ju Ran
通讯作者:Rui Hong;Yanli Xiao;Jie Hu;A. Ji;Shi-Ju Ran
DOI:10.1088/0256-307x/39/10/100701
发表时间:2022-07
期刊:Chinese Physics Letters
影响因子:3.5
作者:Shengxing Bai;Yifan Tang;Shi-Ju Ran
通讯作者:Shengxing Bai;Yifan Tang;Shi-Ju Ran
国内基金
海外基金
