Bose-Einstein condensation (BEC) of triplet excitations triggered by a magnetic field, sometimes called magnon BEC, in dimerized antiferromagnets gives rise to a long-range antiferromagnetic order in the plane perpendicular to the applied magnetic fi
eld. To explore the effects of spin-orbit coupling on magnon condensation, we study the spin model on the distorted honeycomb lattice with dimerized Heisenberg exchange ($J$ terms) and uniform off-diagonal exchange ($Gamma$ terms) interactions. By using variational Monte Carlo method and spin wave theory, we find that an out-of-plane magnetic field can induce different types of long-range magnetic orders, no matter if the ground state is a non-magnetic dimerized state or an antiferromagnetically ordered N{e}el state. Furthermore, the critical properties of field-driven phase transitions in systems with spin-orbit coupling can be different from the conventional magnon BEC. Our study is helpful to understand the rich phases of spin-orbit coupled antiferromagnets in an external magnetic field.
We propose a novel visual memory network architecture for the learning and inference problem in the spatial-temporal domain. Different from the popular transformers, we maintain a fixed set of memory slots in our memory network and explore designs to
input new information into the memory, combine the information in different memory slots and decide when to discard old memory slots. Finally, this architecture is benchmarked on the video object segmentation and video prediction problems. Through the experiments, we show that our memory architecture can achieve competitive results with state-of-the-art while maintaining constant memory capacity.
Light carrying orbital angular momentum constitutes an important resource for both classical and quantum information technologies. Its inherently unbounded nature can be exploited to generate high-dimensional quantum states or for channel multiplexin
g in classical and quantum communication in order to significantly boost the data capacity and the secret key rate, respectively. While the big potentials of light owning orbital angular momentum have been widely ascertained, its technological deployment is still limited by the difficulties deriving from the fabrication of integrated and scalable photonic devices able to generate and manipulate it. Here, we present a photonic integrated chip able to excite orbital angular momentum modes in an 800 m long ring-core fiber, allowing us to perform parallel quantum key distribution using 2 and 3 different modes simultaneously. The experiment sets the first steps towards quantum orbital angular momentum division multiplexing enabled by a compact and light-weight silicon chip, and further pushes the development of integrated scalable devices supporting orbital angular momentum modes.
Despite the significant advances in life science, it still takes decades to translate a basic drug discovery into a cure for human disease. To accelerate the process from bench to bedside, interdisciplinary research (especially research involving bot
h basic research and clinical research) has been strongly recommend by many previous studies. However, the patterns and the roles of the interdisciplinary characteristics in drug research have not been deeply examined in extant studies. The purpose of this study was to characterize interdisciplinary characteristics in drug research from the perspective of translational science, and to examine the role of different kinds of interdisciplinary characteristics in translational research for drugs.
This paper proposes and demonstrates a PHY-layer design of a real-time prototype that supports Ultra-Reliable Communication (URC) in wireless infrastructure networks. The design makes use of Orthogonal Frequency Division Multiple Access (OFDMA) as a
means to achieve URC. Compared with Time-Division Multiple Access (TDMA), OFDMA concentrates the transmit power to a narrower bandwidth, resulting in higher effective SNR. Compared with Frequency-Division Multiple Access (FDMA), OFDMA has higher spectrum efficiency thanks to the smaller subcarrier spacing. Although OFDMA has been introduced in 802.11ax, the purpose was to add flexibility in spectrum usage. Our Reliable OFDMA design, referred to as ROFA, is a clean-slate design with a single goal of ultra-reliable packet delivery. ROFA solves a number of key challenges to ensure the ultra-reliability: (1) a downlink-coordinated time-synchronization mechanism to synchronize the uplink transmission of users, with at most $0.1us$ timing offset; (2) an STF-free packet reception synchronization method that makes use of the property of synchronous systems to avoid packet misdetection; and (3) an uplink precoding mechanism to reduce the CFOs between users and the AP to a negligible level. We implemented ROFA on the Universal Software Radio Peripheral (USRP) SDR platform with real-time signal processing. Extensive experimental results show that ROFA can achieve ultra-reliable packet delivery ($PER<10^5$) with $11.5dB$ less transmit power compared with OFDM-TDMA when they use $3$ and $52$ subcarriers respectively.
In this paper, we establish the second Bogolyubov theorem and global averaging principle for stochastic partial differential equations (in short, SPDEs) with monotone coefficients. Firstly, we prove that there exists a unique $L^{2}$-bounded solution
to SPDEs with monotone coefficients and this bounded solution is globally asymptotically stable in square-mean sense. Then we show that the $L^{2}$-bounded solution possesses the same recurrent properties (e.g. periodic, quasi-periodic, almost periodic, almost automorphic, Birkhoff recurrent, Levitan almost periodic, etc.) in distribution sense as the coefficients. Thirdly, we prove that the recurrent solution of the original equation converges to the stationary solution of averaged equation under the compact-open topology as the time scale goes to zero--in other words, there exists a unique recurrent solution to the original equation in a neighborhood of the stationary solution of averaged equation when the time scale is small. Finally, we establish the global averaging principle in weak sense, i.e. we show that the attractor of original system tends to that of the averaged equation in probability measure space as the time scale goes to zero. For illustration of our results, we give two applications, including stochastic reaction diffusion equations and stochastic generalized porous media equations.
Data augmentation for cross-lingual NER requires fine-grained control over token labels of the augmented text. Existing augmentation approach based on masked language modeling may replace a labeled entity with words of a different class, which makes
the augmented sentence incompatible with the original label sequence, and thus hurts the performance.We propose a data augmentation framework with Masked-Entity Language Modeling (MELM) which effectively ensures the replacing entities fit the original labels. Specifically, MELM linearizes NER labels into sentence context, and thus the fine-tuned MELM is able to predict masked tokens by explicitly conditioning on their labels. Our MELM is agnostic to the source of data to be augmented. Specifically, when MELM is applied to augment training data of the source language, it achieves up to 3.5% F1 score improvement for cross-lingual NER. When unlabeled target data is available and MELM can be further applied to augment pseudo-labeled target data, the performance gain reaches 5.7%. Moreover, MELM consistently outperforms multiple baseline methods for data augmentation.
We are concerned with the uniform regularity estimates of solutions to the two dimensional compressible non-resistive magnetohydrodynamics (MHD) equations with the no-slip boundary condition on velocity in the half plane. Under the assumption that th
e initial magnetic field is transverse to the boundary, the uniform conormal energy estimates are established for the solutions to compressible MHD equations with respect to small viscosity coefficients. As a direct consequence, we proved the inviscid limit of solutions from viscous MHD systems to the ideal MHD systems in $L^infty$ sense. It shows that the transverse magnetic field can prevent the boundary layers from occurring in some physical regime.
Access to resources by users may need to be granted only upon certain conditions and contexts, perhaps particularly in cyber-physical settings. Unfortunately, creating and modifying context-sensitive access control solutions in dynamic environments c
reates ongoing challenges to manage the authorization contexts. This paper proposes RASA, a context-sensitive access authorization approach and mechanism leveraging unsupervised machine learning to automatically infer risk-based authorization decision boundaries. We explore RASA in a healthcare usage environment, wherein cyber and physical conditions create context-specific risks for protecting private health information. The risk levels are associated with access control decisions recommended by a security policy. A coupling method is introduced to track coexistence of the objects within context using frequency and duration of coexistence, and these are clustered to reveal sets of actions with common risk levels; these are used to create authorization decision boundaries. In addition, we propose a method for assessing the risk level and labelling the clusters with respect to their corresponding risk levels. We evaluate the promise of RASA-generated policies against a heuristic rule-based policy. By employing three different coupling features (frequency-based, duration-based, and combined features), the decisions of the unsupervised method and that of the policy are more than 99% consistent.
Sampling is a critical operation in the training of Graph Neural Network (GNN) that helps reduce the cost. Previous works have explored improving sampling algorithms through mathematical and statistical methods. However, there is a gap between sampli
ng algorithms and hardware. Without consideration of hardware, algorithm designers merely optimize sampling at the algorithm level, missing the great potential of promoting the efficiency of existing sampling algorithms by leveraging hardware features. In this paper, we first propose a unified programming model for mainstream sampling algorithms, termed GNNSampler, covering the key processes for sampling algorithms in various categories. Second, we explore the data locality among nodes and their neighbors (i.e., the hardware feature) in real-world datasets for alleviating the irregular memory access in sampling. Third, we implement locality-aware optimizations in GNNSampler for diverse sampling algorithms to optimize the general sampling process in the training of GNN. Finally, we emphatically conduct experiments on large graph datasets to analyze the relevance between the training time, model accuracy, and hardware-level metrics, which helps achieve a good trade-off between time and accuracy in GNN training. Extensive experimental results show that our method is universal to mainstream sampling algorithms and reduces the training time of GNN (range from 4.83% with layer-wise sampling to 44.92% with subgraph-based sampling) with comparable accuracy.
