arXiv subjects

Gang Chen

Publications and source records attributed to Gang Chen.

At least 361 records · Page 20Linked to original sources

arXiv2021

3D hinge transport in acoustic higher-order topological insulators

The discovery of topologically protected boundary states in topological insulators opens a new avenue toward exploring novel transport phenomena. The one-way feature of boundary states against disorders and impurities prospects great potential in applications of electronic and classical wave devices. Particularly, for the 3D higher-order topological insulators, it can host hinge states, which allow the energy to transport along the hinge channels. However, the hinge states haveonly been observed along a single hinge, and a natural question arises: whether the hinge states can exist simultaneously on all the three independent directions of one sample? Here we theoretically predict and experimentally observe the hinge states on three different directions of a higher-order topological phononic crystal, and demonstrate their robust one-way transport from hinge to hinge. Therefore, 3D topological hinge transport is successfully achieved. The novel sound transport may serve as the basis for acoustic devices of unconventional functions.

cond-mat.mes-hall↗

arXiv2021

Evaluation of diffuse mismatch model for phonon scattering at disordered interfaces

Diffuse phonon scattering strongly affects the phonon transport through a disordered interface. The often-used diffuse mismatch model assumes that phonons lose memory of their origin after being scattered by the interface. Using mode-resolved atomic Green's function simulation, we demonstrate that diffuse phonon scattering by a single disordered interface cannot make a phonon lose its memory and thus the applicability of diffusive mismatch model is limited. An analytical expression for diffuse scattering probability based on the continuum approximation is also derived and shown to work reasonably well at low frequencies.

cond-mat.mes-hall↗

arXiv2021

Twisted magnetic topological insulators

Motivated by the discovery of the quantum anomalous Hall effect in Cr-doped \ce{(Bi,Sb)2Te3} thin films, we study the generic states for magnetic topological insulators and explore the physical properties for both magnetism and itinerant electrons. First-principles calculations are exploited to investigate the magnetic interactions between magnetic Co atoms adsorbed on the \ce{Bi2Se3} (111) surface. Due to the absence of inversion symmetry on the surface, there are Dzyaloshinskii-Moriya-like twisted spin interactions between the local moments of Co ions. These nonferromagnetic interactions twist the collinear spin configuration of the ferromagnet and generate various magnetic orders beyond a simple ferromagnet. Among them, the spin spiral state generates alternating counterpropagating modes across each period of spin states, and the skyrmion lattice even supports a chiral mode around the core of each skyrmion. The skyrmion lattice opens a gap at the surface Dirac point, resulting in the anomalous Hall effect. These results may inspire further experimental investigation of magnetic topological insulators.

cond-mat.str-el↗

arXiv2021

ARM-Net: Adaptive Relation Modeling Network for Structured Data

Relational databases are the de facto standard for storing and querying structured data, and extracting insights from structured data requires advanced analytics. Deep neural networks (DNNs) have achieved super-human prediction performance in particular data types, e.g., images. However, existing DNNs may not produce meaningful results when applied to structured data. The reason is that there are correlations and dependencies across combinations of attribute values in a table, and these do not follow simple additive patterns that can be easily mimicked by a DNN. The number of possible such cross features is combinatorial, making them computationally prohibitive to model. Furthermore, the deployment of learning models in real-world applications has also highlighted the need for interpretability, especially for high-stakes applications, which remains another issue of concern to DNNs. In this paper, we present ARM-Net, an adaptive relation modeling network tailored for structured data, and a lightweight framework ARMOR based on ARM-Net for relational data analytics. The key idea is to model feature interactions with cross features selectively and dynamically, by first transforming the input features into exponential space, and then determining the interaction order and interaction weights adaptively for each cross feature. We propose a novel sparse attention mechanism to dynamically generate the interaction weights given the input tuple, so that we can explicitly model cross features of arbitrary orders with noisy features filtered selectively. Then during model inference, ARM-Net can specify the cross features being used for each prediction for higher accuracy and better interpretability. Our extensive experiments on real-world datasets demonstrate that ARM-Net consistently outperforms existing models and provides more interpretable predictions for data-driven decision making.

cs.LG↗

arXiv2021

A new gauge-invariant double copy for heavy-mass effective theory

We propose a new form of the colour-kinematics/double-copy duality for heavy-mass effective field theories, which we apply to construct compact expressions for tree amplitudes with heavy matter particles in Yang-Mills and in gravity to leading order in the mass. In this set-up, the new BCJ numerators are fixed uniquely and directly written in terms of field strengths, making gauge invariance manifest. Furthermore, they are local and automatically satisfy the Jacobi relations and crossing symmetry. We construct these BCJ numerators explicitly up to six particles. We also discuss relations of the BCJ numerators in the heavy-mass effective theory with those in pure Yang-Mills amplitudes.

hep-th↗

arXiv2021

Joining datasets via data augmentation in the label space for neural networks

Most, if not all, modern deep learning systems restrict themselves to a single dataset for neural network training and inference. In this article, we are interested in systematic ways to join datasets that are made of similar purposes. Unlike previous published works that ubiquitously conduct the dataset joining in the uninterpretable latent vectorial space, the core to our method is an augmentation procedure in the label space. The primary challenge to address the label space for dataset joining is the discrepancy between labels: non-overlapping label annotation sets, different labeling granularity or hierarchy and etc. Notably we propose a new technique leveraging artificially created knowledge graph, recurrent neural networks and policy gradient that successfully achieve the dataset joining in the label space. Empirical results on both image and text classification justify the validity of our approach.

cs.LG↗

arXiv2021

A critical look at the current train/test split in machine learning

The randomized or cross-validated split of training and testing sets has been adopted as the gold standard of machine learning for decades. The establishment of these split protocols are based on two assumptions: (i)-fixing the dataset to be eternally static so we could evaluate different machine learning algorithms or models; (ii)-there is a complete set of annotated data available to researchers or industrial practitioners. However, in this article, we intend to take a closer and critical look at the split protocol itself and point out its weakness and limitation, especially for industrial applications. In many real-world problems, we must acknowledge that there are numerous situations where assumption (ii) does not hold. For instance, for interdisciplinary applications like drug discovery, it often requires real lab experiments to annotate data which poses huge costs in both time and financial considerations. In other words, it can be very difficult or even impossible to satisfy assumption (ii). In this article, we intend to access this problem and reiterate the paradigm of active learning, and investigate its potential on solving problems under unconventional train/test split protocols. We further propose a new adaptive active learning architecture (AAL) which involves an adaptation policy, in comparison with the traditional active learning that only unidirectionally adds data points to the training pool. We primarily justify our points by extensively investigating an interdisciplinary drug-protein binding problem. We additionally evaluate AAL on more conventional machine learning benchmarking datasets like CIFAR-10 to demonstrate the generalizability and efficacy of the new framework.

cs.LG↗

arXiv2021

Generic Spiral Spin Liquids

Spiral spin liquids are unique classical spin liquids that occur in many frustrated spin systems, but do not comprise a new phase of matter. Owing to extensive classical ground-state degeneracy, the spins in a spiral spin liquid thermally fluctuate cooperatively from a collection of spiral configurations at low temperatures. These spiral propagation wavevectors form a continuous manifold in reciprocal space, \textit{i.e.}, a spiral contour or a spiral surface, that strongly governs the low-temperature thermal fluctuations and magnetic physics. In this paper, the relevant spin models conveying the spiral spin liquid physics are systematically explored and the geometric origin of the spiral manifold is clarified in the model construction. The spiral spin liquids based on the dimension and the codimension of the spiral manifold are further clarified. For each class, the physical properties are studied both generally and for specific examples. The results are relevant to a wide range of frustrated magnets. A survey of materials is given and future experiments are suggested.

cond-mat.str-el↗

arXiv2021

Topological chiral spin liquids and competing states in triangular lattice SU($N$) Mott insulators

SU($N$) Mott insulators have been proposed and/or realized in solid-state materials and with ultracold atoms on optical lattices. We study the two-dimensional SU($N$) antiferromagnets on the triangular lattice. Starting from an SU($N$) Heisenberg model with the fundamental representation on each site in the large-$N$ limit, we perform a self-consistent calculation and find a variety of ground states including the valence cluster states, stripe ordered states with a doubled unit-cell and topological chiral spin liquids. The system favors a cluster or ordered ground state when the number of flavors $N$ is less than 6. It is shown that, increasing the number of flavors enhances quantum fluctuations and eventually transfer the clusterized ground states into a topological chiral spin liquids. This chiral spin liquid ground state has an equivalent for the square lattice SU($N$) magnets. We further identify the corresponding lowest competing states that represent another distinct type of chiral spin liquid states. We conclude with a discussion about the relevant systems and the experimental probes.

cond-mat.quant-gas↗

arXiv2021

Tensor products of coherent configurations

A Cartesian decomposition of a coherent configuration $\cal X$ is defined as a special set of its parabolics that form a Cartesian decomposition of the underlying set. It turns out that every tensor decomposition of $\cal X$ comes from a certain Cartesian decomposition. It is proved that if the coherent configuration $\cal X$ is thick, then there is a unique maximal Cartesian decomposition of $\cal X$, i.e., there is exactly one internal tensor decomposition of $\cal X$ into indecomposable components. In particular, this implies an analog of the Krull--Schmidt theorem for the thick coherent configurations. A polynomial-time algorithm for finding the maximal Cartesian decomposition of a thick coherent configuration is constructed.

math.CO↗

arXiv2021

Learning Symbolic Expressions via Gumbel-Max Equation Learner Networks

Most of the neural networks (NNs) learned via state-of-the-art machine learning techniques are black-box models. For a widespread success of machine learning in science and engineering, it is important to develop new NN architectures to effectively extract high-level mathematical knowledge from complex datasets. Motivated by this understanding, this paper develops a new NN architecture called the Gumbel-Max Equation Learner (GMEQL) network. Different from previously proposed Equation Learner (EQL) networks, GMEQL applies continuous relaxation to the network structure via the Gumbel-Max trick and introduces two types of trainable parameters: structure parameters and regression parameters. This paper also proposes a two-stage training process with new techniques to train structure parameters in both online and offline settings based on an elite repository. On 8 benchmark symbolic regression problems, GMEQL is experimentally shown to outperform several cutting-edge machine learning approaches.

cs.LG↗

arXiv2021

Next-to-MHV Yang-Mills kinematic algebra

Kinematic numerators of Yang-Mills scattering amplitudes possess a rich Lie algebraic structure that suggest the existence of a hidden infinite-dimensional kinematic algebra. Explicitly realizing such a kinematic algebra is a longstanding open problem that only has had partial success for simple helicity sectors. In past work, we introduced a framework using tensor currents and fusion rules to generate BCJ numerators of a special subsector of NMHV amplitudes in Yang-Mills theory. Here we enlarge the scope and explicitly realize a kinematic algebra for all NMHV amplitudes. Master numerators are obtained directly from the algebraic rules and through commutators and kinematic Jacobi identities other numerators can be generated. Inspecting the output of the algebra, we conjecture a closed-form expression for the master BCJ numerator up to any multiplicity. We also introduce a new method, based on group algebra of the permutation group, to solve for the generalized gauge freedom of BCJ numerators. It uses the recently introduced binary BCJ relations to provide a complete set of NMHV kinematic numerators that consist of pure gauge.

hep-th↗

arXiv2021

Machine vision detection to daily facial fatigue with a nonlocal 3D attention network

Fatigue detection is valued for people to keep mental health and prevent safety accidents. However, detecting facial fatigue, especially mild fatigue in the real world via machine vision is still a challenging issue due to lack of non-lab dataset and well-defined algorithms. In order to improve the detection capability on facial fatigue that can be used widely in daily life, this paper provided an audiovisual dataset named DLFD (daily-life fatigue dataset) which reflected people's facial fatigue state in the wild. A framework using 3D-ResNet along with non-local attention mechanism was training for extraction of local and long-range features in spatial and temporal dimensions. Then, a compacted loss function combining mean squared error and cross-entropy was designed to predict both continuous and categorical fatigue degrees. Our proposed framework has reached an average accuracy of 90.8% on validation set and 72.5% on test set for binary classification, standing a good position compared to other state-of-the-art methods. The analysis of feature map visualization revealed that our framework captured facial dynamics and attempted to build a connection with fatigue state. Our experimental results in multiple metrics proved that our framework captured some typical, micro and dynamic facial features along spatiotemporal dimensions, contributing to the mild fatigue detection in the wild.

cs.CV↗

arXiv2021

Model Slicing for Supporting Complex Analytics with Elastic Inference Cost and Resource Constraints

Deep learning models have been used to support analytics beyond simple aggregation, where deeper and wider models have been shown to yield great results. These models consume a huge amount of memory and computational operations. However, most of the large-scale industrial applications are often computational budget constrained. In practice, the peak workload of inference service could be 10x higher than the average cases, with the presence of unpredictable extreme cases. Lots of computational resources could be wasted during off-peak hours and the system may crash when the workload exceeds system capacity. How to support deep learning services with a dynamic workload cost-efficiently remains a challenging problem. In this paper, we address the challenge with a general and novel training scheme called model slicing, which enables deep learning models to provide predictions within the prescribed computational resource budget dynamically. Model slicing could be viewed as an elastic computation solution without requiring more computational resources. Succinctly, each layer in the model is divided into groups of a contiguous block of basic components (i.e. neurons in dense layers and channels in convolutional layers), and then partially ordered relation is introduced to these groups by enforcing that groups participated in each forward pass always starts from the first group to the dynamically-determined rightmost group. Trained by dynamically indexing the rightmost group with a single parameter slice rate, the network is engendered to build up group-wise and residual representation. Then during inference, a sub-model with fewer groups can be readily deployed for efficiency whose computation is roughly quadratic to the width controlled by the slice rate. Extensive experiments show that models trained with model slicing can effectively support on-demand workload with elastic inference cost.

cs.LG↗

arXiv2021

Towards a Universal Continuous Knowledge Base

In artificial intelligence (AI), knowledge is the information required by an intelligent system to accomplish tasks. While traditional knowledge bases use discrete, symbolic representations, detecting knowledge encoded in the continuous representations learned from data has received increasing attention recently. In this work, we propose a method for building a continuous knowledge base (CKB) that can store knowledge imported from multiple, diverse neural networks. The key idea of our approach is to define an interface for each neural network and cast knowledge transferring as a function simulation problem. Experiments on text classification show promising results: the CKB imports knowledge from a single model and then exports the knowledge to a new model, achieving comparable performance with the original model. More interesting, we import the knowledge from multiple models to the knowledge base, from which the fused knowledge is exported back to a single model, achieving a higher accuracy than the original model. With the CKB, it is also easy to achieve knowledge distillation and transfer learning. Our work opens the door to building a universal continuous knowledge base to collect, store, and organize all continuous knowledge encoded in various neural networks trained for different AI tasks.

cs.CL↗

arXiv2021

Ferromagnetism in 2D Vanadium Diselenide

Two-dimensional (2D) Van der Waals ferromagnets carry the promise of ultimately miniature spintronics and information storage devices. Among the newly discovered 2D ferromagnets all inherit the magnetic ordering from their bulk ancestors. Here we report a new 2D ferromagnetic semiconductor at room temperature, 2H phase vanadium diselenide (VSe2) which show ferromagnetic at 2D form only. This unique 2D ferromagnetic semiconductor manifests an enhanced magnetic ordering owing to structural anisotropy at 2D limit.

cond-mat.mtrl-sci↗

arXiv2021

Writing and deleting skyrmions with electric fields in a multiferroic heterostructure

Magnetic skyrmions are topological spin textures that can be used as information carriers for the next-generation information storage and processing. The electric-field controlling of skyrmions in such devices is essential but remains technologically challenging. Here, using the first-principles calculation and the Ginzburg-Landau theory, we propose a reliable process for writing and deleting skyrmions by electric fields, on the platform of a multiferroic heterostructure, particularly the $\text{Cr}_{2}\text{Ge}_{2}\text{Te}_{6} $/$ \text{In}_{2}\text{Se}_{3} $ heterostructure. We show that the electric field controls the electric polarization and indirectly influences the antisymmetric Dzyaloshinskii-Moriya interaction (DMI) between the magnetic moments. The latter is responsible for the generation and removal of the skyrmion spin textures, and we study this mechanism by the Ginzburg-Landau analysis. We discuss the real-space Berry curvature, topological Hall effects, possible quantum anomalous Hall effect, and other competing magnetic structures. These results represent examples of quantum technology and may have potential applications in future skyrmionics and the device fabrication.

cond-mat.str-el↗

arXiv2021

Spinon Fermi surface spin liquid in a triangular lattice antiferromagnet NaYbSe$_2$

Triangular lattice of rare-earth ions with interacting effective spin-$1/2$ local moments is an ideal platform to explore the physics of quantum spin liquids (QSLs) in the presence of strong spin-orbit coupling, crystal electric fields, and geometrical frustration. The Yb delafossites, NaYbCh$_2$ (Ch=O, S, Se) with Yb ions forming a perfect triangular lattice, have been suggested to be candidates for QSLs. Previous thermodynamics, nuclear magnetic resonance, and muon spin rotation measurements on NaYbCh$_2$ have supported the suggestion of the QSL ground states. The key signature of a QSL, the spin excitation continuum, arising from the spin quantum number fractionalization, has not been observed. Here we perform both elastic and inelastic neutron scattering measurements as well as detailed thermodynamic measurements on high-quality single-crystalline NaYbSe$_2$ samples to confirm the absence of long-range magnetic order down to 40 mK, and further reveal a clear signature of magnetic excitation continuum extending from 0.1 to 2.5 meV. The comparison between the structure of the magnetic excitation spectra and the theoretical expectation from the spinon continuum suggests that the ground state of NaYbSe$_2$ is a QSL with a spinon Fermi surface.

cond-mat.str-el↗