神经网络量子态及其应用

神经网络量子态是由人工神经网络所表示的量子态。得益于机器学习,尤其是深度学习近年来取得的突破性进展,神经网络量子态的研究得到了广泛的关注,成为当前的热点前沿方向。文章将介绍不同的神经网络量子态,其物理性质与典型应用场景,最新进展,以及面临的挑战。     

Probe Knots and Hopf Insulators with Ultracold Atoms

Knots and links are fascinating and intricate topological objects. Their influence spans from DNA and molecular chemistry to vortices in superfluid helium, defects in liquid crystals and cosmic strings in the early universe. Here we find that knotted structures also exist in a peculiar class of three-dimensional topological insulators—the Hopf insulators. In particular, we demonstrate that the momentum-space spin textures of Hopf insulators are twisted in a nontrivial way, which implies the presence of various knot and link structures. We further illustrate that the knots and nontrivial spin textures can be probed via standard time-of-flight images in cold atoms as preimage contours of spin orientations in stereographic coordinates. The extracted Hopf invariants, knots, and links are validated to be robust to typical experimental imperfections. Our work establishes the existence of knotted structures in Hopf insulators, which may have potential applications in spintronics and quantum information processing.     

数字化量子模拟拓扑时间晶体

<正>处于非平衡态的量子多体系统中存在许多新奇的物理现象。对称、拓扑以及局域化等思想的引入,可以进一步丰富这类系统的相图^[1—3]。如何进一步深化对这些非平衡物质相的理解是一个广为人知的科学挑战：无论是在理论还是在数值方面,我们都缺乏有力工具来深入地研究这些非平衡物相及相变过程。最近几年,随着量子计算和量子模拟技术的迅速发展,人们期待这些人工量子系统可以成为探究多体量子物理的有力工具。     

量子人工智能中的对抗学习

量子人工智能是一个探究人工智能与量子物理交叉的领域:一方面人工智能的方法和技术可以用来解决量子科学中的问题;另一方面,量子计算的发展也可能为人工智能,尤其是机器学习,提供新的范式,极大促进人工智能的发展.然而,量子机器学习和经典学习系统对于对抗样本同样具有脆弱性:在原始数据样本上添加精心制作的微小扰动将很可能导致系统做出错误的预测.本文介绍经典与量子对抗机器学习的基本概念、原理、以及最新进展.首先从经典和量子两个方面介绍对抗学习,通过二维经典伊辛模型和三维手征拓扑绝缘体的对抗样本揭示出经典机器学习在识别物质相时的脆弱性,同时利用手写字体的对抗样本直观展示出量子分类器的脆弱性.随后从理论层面上分别阐述经典与量子的"没有免费午餐"定理,并探讨了量子分类器的普适对抗样本.最后,分析并讨论了相应的防御策略.量子人工智能中对抗学习的研究揭示了量子智能系统潜在的风险以及可能的防御策略,将对未来量子技术与人工智能的交叉产生深刻影响.     

Quantum Continual Learning Overcoming Catastrophic Forgetting

Catastrophic forgetting describes the fact that machine learning models will likely forget the knowledge of previously learned tasks after the learning process of a new one.It is a vital problem in the continual learning scenario and recently has attracted tremendous concern across different communities.We explore the catastrophic forgetting phenomena in the context of quantum machine learning.It is found that,similar to those classical learning models based on neural networks,quantum learning s...     

Proposed Entanglement Swapping in Continuous Variable Systems via Braiding

We study entanglement swapping in continuous variable systems by using braiding transformations. It is found that entanglement swapping in two-mode squeezed vacuum states and squeezed coherent states can be realized based on the braiding operators.     

四价掺杂铌酸锂晶体抗光折变性能研究

我们在同成份铌酸锂晶体中掺入四价离子铪,生长了掺杂浓度分别为2、4、6mol;的掺铪铌酸锂系列晶体.掺铪浓度达到4mol;时,晶体的抗光损伤能力为5×105W/cm2,比同成份纯铌酸锂晶体提高了4个数量级.应用全息法测得掺4、6mol;铪的铌酸锂晶体最大折射率变化为8.7×10-6,与高掺镁(6.5mol;)铌酸锂晶体的类似.晶体的红外吸收谱和紫外-可见光吸收谱也显示,掺杂浓度为4mol;时具有明显的阈值特征.由此可以确定铪离子在铌酸锂晶体中的阈值浓度约为4mol;.