Review
Nanoscience & Nanotechnology
Laura Pilozzi, Francis A. Farrelly, Giulia Marcucci, Claudio Conti
Summary: This study proposes the use of artificial neural networks to design and characterize photonic topological insulators, aiming to obtain protected edge states at target frequencies. By applying machine learning, one can identify the parameters of topological insulators and solve long-standing inverse problems.
Article
Computer Science, Artificial Intelligence
Oleksandr Balabanov, Mats Granath
Summary: This study introduces a neural-network-based protocol for identifying topologically relevant indices invariant under transformations between trivial atomic-limit Hamiltonians, thus enabling standard classification of band insulators. The work extends the method of 'topological data augmentation' by generalizing and simplifying the data generation scheme and introducing a special 'mod' layer in the neural network suitable for Z(n) classification.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2021)
Article
Chemistry, Physical
Agnese Marcato, Javier E. Santos, Chaoyue Liu, Gianluca Boccardo, Daniele Marchisio, Alejandro A. Franco
Summary: The article introduces the development trend of investigating physical phenomena in the 3D porous microstructure of electrodes in the field of lithium-ion batteries. It proposes a multiscale network model based on machine learning to predict the discharge behavior of lithium-ion batteries.
ENERGY STORAGE MATERIALS
(2023)
Review
Chemistry, Multidisciplinary
Zhaocheng Liu, Dayu Zhu, Lakshmi Raju, Wenshan Cai
Summary: Machine learning, as an algorithm study for automated prediction and decision-making based on complex data, has become an indispensable tool in artificial intelligence research, leading to significant progress in scientific research fields like quantum physics, organic chemistry, and medical imaging. Recently adopted in photonics and optics research, machine learning shows potential for addressing the inverse design problem and has been instrumental in advancing photonic design strategies in recent years, particularly focusing on deep learning methods for high degrees-of-freedom structure design.
Article
Management
Marta Monaci, Valerio Agasucci, Giorgio Grani
Summary: In this research, we applied deep reinforcement learning to tackle the job shop scheduling problem. The study showed that a greedy-like heuristic trained on a subset of problems could effectively generalize to unseen instances and be competitive compared to other methods. The experiments demonstrated that this algorithm was able to generate good solutions in a short time, indicating the feasibility of learning-based methodologies in generating new greedy heuristics.
EUROPEAN JOURNAL OF OPERATIONAL RESEARCH
(2024)
Article
Engineering, Mechanical
Tianju Xue, Sigrid Adriaenssens, Sheng Mao
Summary: This research proposes a machine learning approach, called "Metamaterial Graph Network (MGN)", to predict the nonlinear dynamic response of soft mechanical metamaterials with complex geometries. MGN is capable of simulating large-scale metamaterial structures and successfully captures their unique wave propagation behavior.
INTERNATIONAL JOURNAL OF MECHANICAL SCIENCES
(2023)
Article
Computer Science, Artificial Intelligence
Yuchen Zhang, Jizhe Liu, Yan Xu, Zhao Yang Dong
Summary: Constrained optimization plays a crucial role in solving real-world problems. Existing data-driven approaches for constrained optimization only work for simple constraints. This paper proposes an athlete-referee dual learning system for end-to-end constrained optimization with large-scale complex constraints.
KNOWLEDGE-BASED SYSTEMS
(2023)
Article
Geosciences, Multidisciplinary
B. Balogh, D. Saint-Martin, A. Ribes
Summary: This article proposes a method to design tunable machine learning parameterizations and calibrate them online. By combining model parameters with machine learning model inputs, a range of parameter values are fitted at once using an offline metric, and the parameters included in the machine learning inputs are optimized with respect to an online metric to reduce long-term biases of the machine learning model.
GEOPHYSICAL RESEARCH LETTERS
(2022)
Article
Computer Science, Artificial Intelligence
Luke Kljucaric, Alan D. George
Summary: As computer architectures integrate application-specific hardware, understanding the relative performance of devices becomes crucial. Benchmarking suites like MLPerf aim to standardize fair comparisons of different hardware architectures, but many apps require different workloads and have specific performance goals. This research analyzes compute architectures for handwritten Chinese character recognition, comparing latency and throughput for different models using ML-specific hardware.
ACM TRANSACTIONS ON INTELLIGENT SYSTEMS AND TECHNOLOGY
(2023)
Article
Computer Science, Artificial Intelligence
Li Li, Duo Liu, Moming Duan, Yu Zhang, Ao Ren, Xianzhang Chen, Yujuan Tan, Chengliang Wang
Summary: In this paper, a novel adaptive federated learning framework is proposed to address the challenges of system heterogeneity and improve the robustness of the global model. By leveraging the workload completion history, the framework can predict the affordable training workload for each device, reducing stragglers in highly heterogeneous systems. Active learning is incorporated to dynamically schedule participants, accelerating model convergence. Experimental evaluations demonstrate the effectiveness of the framework compared to other methods.
Article
Nanoscience & Nanotechnology
Fengrui Wang, Shiyi Qin, Claribel Acevedo-Velez, Reid C. Van Lehn, Victor M. Zavala, David M. Lynn
Summary: Reported is a surfactant sensing platform based on liquid crystals and machine learning, which can reliably predict the concentration and type of surfactants in aqueous solutions and provide computational capabilities for the optical features of droplets.
ACS APPLIED MATERIALS & INTERFACES
(2023)
Article
Computer Science, Theory & Methods
William C. Sleeman, Rishabh Kapoor, Preetam Ghosh
Summary: This study proposes a new taxonomy for describing multimodal classification models, aiming to address challenges in the field such as inconsistent terminologies and architectural descriptions, as well as unresolved issues like big data, class imbalance, and instance-level difficulty. The paper presents examples of applying this taxonomy to existing models and offers a checklist for the clear and complete presentation of future models.
ACM COMPUTING SURVEYS
(2023)
Article
Computer Science, Artificial Intelligence
Krishnan Raghavan, Shweta Garg, Sarangapani Jagannathan, V. A. Samaranayake
Summary: This article introduces a novel learning methodology for classification in high-dimensional data, addressing challenges through a L-1 regularized zero-sum game. The proposed approach utilizes distributed learning and an alternating minimization method for optimal sparsity. The efficiency of the approach is demonstrated theoretically and empirically with nine data sets.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2021)
Article
Nanoscience & Nanotechnology
Mohammad-Ali Miri, Vinod Menon
Summary: We demonstrate that coherent laser networks (CLNs) possess emerging neural computing capabilities by utilizing their collective behavior for storing and retrieving phase patterns. Moreover, the limitation of pattern retrieval to only two images can be overcome by introducing nonreciprocal coupling between lasers, thus enabling a larger storage capacity. This research opens up new possibilities for neural computation using coherent laser networks as analog processors, and introduces a novel energy-based recurrent neural network that handles continuous data.
Article
Computer Science, Theory & Methods
Moming Duan, Duo Liu, Xianzhang Chen, Renping Liu, Yujuan Tan, Liang Liang
Summary: Federated Learning (FL) is a distributed deep learning method where multiple devices contribute to a neural network training while keeping their data private. Data imbalance in mobile systems can lead to accuracy degradation in FL applications, but the Astraea framework offers improvements through data augmentation and rescheduling. Compared to FedAvg, Astraea demonstrates higher accuracy and reduced communication traffic.
IEEE TRANSACTIONS ON PARALLEL AND DISTRIBUTED SYSTEMS
(2021)
Article
Physics, Particles & Fields
Kaixiang Su, Pengfei Zhang, Hui Zhai
Summary: Using the Sachdev-Ye-Kitaev model, this paper examines the entropy dynamics of a quantum system interacting with a non-Markovian environment, finding that entropy initially increases linearly but decreases later due to non-Markovian effects, forming a Page curve. The exact solution of this model is believed to be universally applicable to chaotic quantum many-body systems and can potentially be experimentally verified in the near future.
JOURNAL OF HIGH ENERGY PHYSICS
(2021)
Article
Physics, Multidisciplinary
Zhe-Yu Shi, Chao Gao, Hui Zhai
Summary: This article discusses the mathematical connection between ideal gases and hydrodynamics, exploring situations where the solution to a class of interacting hydrodynamic equations can be constructed from the dynamics of noninteracting ideal gases. Three examples are provided, with implications for experimental observations and future predictions in ultracold atomic gases. The ideal-gas approach to hydrodynamics may have broad applications in different subfields of physics.
Article
Multidisciplinary Sciences
Libo Liang, Wei Zheng, Ruixiao Yao, Qinpei Zheng, Zhiyuan Yao, Tian-Gang Zhou, Qi Huang, Zhongchi Zhang, Jilai Ye, Xiaoji Zhou, Xuzong Chen, Wenlan Chen, Hui Zhai, Jiazhong Hu
Summary: The article presents a novel method of probing quantum many-body correlation by ramping dynamics. The researchers demonstrate this method experimentally by studying the Bose-Hubbard model with ultracold atoms in three-dimensional optical lattices. This method provides important insights into the physical properties of quantum systems.
Article
Quantum Science & Technology
Fan Yang, Hui Zhai
Summary: In this work, a protocol to measure quantized nonlinear transport using ultracold atomic Fermi gases is proposed. Practical effects in experiments are investigated and a method to reduce deviation is proposed based on symmetry considerations. The quantized nonlinear response can be observed reasonably well under experimental conditions readily achieved with ultracold atoms.
Article
Physics, Multidisciplinary
Yanting Cheng, Chengshu Li, Hui Zhai
Summary: Recently, the Rydberg blockade effect has been used to realize quantum spin liquid (QSL) on a kagome lattice, and evidence of QSL has been obtained experimentally by measuring non-local string order. In this paper, a Bardeen-Cooper-Schrieffer (BCS)-type variational wave function study is reported for the spin liquid state in this model, which is motivated by mapping the Rydberg blockade model to a lattice gauge theory. The predictions of this wave function are compared with experimental measurements of non-local string order, and good agreement is found.
NEW JOURNAL OF PHYSICS
(2023)
Article
Physics, Multidisciplinary
Chang Liu, Haifeng Tang, Hui Zhai
Summary: In this paper, the authors generalize Krylov complexity from a closed system to an open system coupled to a Markovian bath, where Lindbladian evolution replaces Hamiltonian evolution. They show that the Krylov complexity in open systems can be mapped to a non-Hermitian tight-binding model in a half-infinite chain. The strength of the non-Hermitian terms increases linearly with the increase of the Krylov basis index n.
PHYSICAL REVIEW RESEARCH
(2023)
Article
Quantum Science & Technology
Yanting Cheng, Shang Liu, Wei Zheng, Pengfei Zhang, Hui Zhai
Summary: This article introduces the realization of the one-dimensional lattice Schwinger model using bosons and Rydberg-atom arrays, and discusses methods to study confinement and deconfinement by varying the mass of the matter field and tuning the topological angle.
Article
Physics, Multidisciplinary
Lei Pan, Hui Zhai
Summary: In this Letter, a composite spin representation is proposed to provide a unified description for the correlated effects in Rydberg atom arrays. The ground state and excitation spectrum can be accurately described using composite spins, and the differences between arrays with different blocking radii can be absorbed in the formation of composite spins. This scheme offers a universal physical picture of the blockade effect in Rydberg atom arrays.
PHYSICAL REVIEW RESEARCH
(2022)
Article
Physics, Multidisciplinary
Tian-Gang Zhou, Yi-Neng Zhou, Pengfei Zhang, Hui Zhai
Summary: This article studies the connection between quantum chaos in Hermitian systems and the skin effect in non-Hermitian systems, and unifies them through the concept of space-time duality.
PHYSICAL REVIEW RESEARCH
(2022)
Article
Materials Science, Multidisciplinary
Zhiyuan Yao, Lei Pan, Shang Liu, Hui Zhai
Summary: In this paper, we investigate the PXP Hamiltonian with an external magnetic field and discover surprising connections between quantum scar states and quantum criticality. We show that the quantum many-body scar states can be traced to quantum critical states and the violation of quantum thermalization diminishes in the quantum critical regime. These findings are important for the verification on existing cold atom experiment platforms.
Article
Computer Science, Artificial Intelligence
Juan Yao, Ce Wang, Zhiyuan Yao, Hui Zhai
Summary: In this work, a neural network-based method is developed to solve the problem of analytic continuation, which is an important bridge between many-body theories and experiments. The trained neural network shows the best performance when a proper amount of noise is added to the training data. The method can successfully capture multi-peak structure in the resulting response function and can be combined with Monte Carlo simulations for comparing with experiments on real-time dynamics.
MACHINE LEARNING-SCIENCE AND TECHNOLOGY
(2022)
Article
Materials Science, Multidisciplinary
Peng Xu, Wei Zheng, Hui Zhai
Summary: The Floquet Hamiltonian is not sufficient to accurately describe a time-periodic system. A set of Hamiltonians spanning all values of the micromotion parameter is needed to provide an accurate description. These micromotion parameters can be seen as an extra dimension of the system, and the topological invariants defined in this higher-dimensional system can ensure the presence of edge states in the lower-dimensional Floquet system.
Article
Physics, Multidisciplinary
Yi-Neng Zhou, Liang Mao, Hui Zhai
Summary: The paper investigates the segment structure and two different energy scales in the Lindblad spectrum of a quantum many-body system in the strong dissipation regime, leading to opposite behaviors in short-time and long-time dynamics. With increasing dissipation strength, short-time dynamics become fast while long-time dynamics become slow, and entropy dynamics are suppressed. This is demonstrated through numerical studies of two experimentally verifiable models.
PHYSICAL REVIEW RESEARCH
(2021)
Article
Physics, Multidisciplinary
Yadong Wu, Pengfei Zhang, Hui Zhai
Summary: By characterizing quantum information scrambling through operator size growth, we introduce an averaged operator size to describe the information scrambling ability of a quantum neural network architecture, suggesting a positive correlation with learning efficiency. Our study on various architectures and learning tasks shows that architectures with larger averaged operator size exhibit higher learning efficiency, indicated by faster decrease in loss function or increase in prediction accuracy as training epochs increase.
PHYSICAL REVIEW RESEARCH
(2021)
Article
Physics, Multidisciplinary
Yadong Wu, Juan Yao, Pengfei Zhang, Hui Zhai
Summary: The study investigates how a deep quantum neural network can approximate a target function as accurately as possible, finding that accuracy is achievable when input wave functions in the dataset do not span the entire Hilbert space.
PHYSICAL REVIEW RESEARCH
(2021)