arXiv subjects

Bin Liu

Publications and source records attributed to Bin Liu.

At least 433 records · Page 24Linked to original sources

arXiv2019

Dynamic Ensemble Modeling Approach to Nonstationary Neural Decoding in Brain-Computer Interfaces

Brain-computer interfaces (BCIs) have enabled prosthetic device control by decoding motor movements from neural activities. Neural signals recorded from cortex exhibit nonstationary property due to abrupt noises and neuroplastic changes in brain activities during motor control. Current state-of-the-art neural signal decoders such as Kalman filter assume fixed relationship between neural activities and motor movements, thus will fail if this assumption is not satisfied. We propose a dynamic ensemble modeling (DyEnsemble) approach that is capable of adapting to changes in neural signals by employing a proper combination of decoding functions. The DyEnsemble method firstly learns a set of diverse candidate models. Then, it dynamically selects and combines these models online according to Bayesian updating mechanism. Our method can mitigate the effect of noises and cope with different task behaviors by automatic model switching, thus gives more accurate predictions. Experiments with neural data demonstrate that the DyEnsemble method outperforms Kalman filters remarkably, and its advantage is more obvious with noisy signals.

eess.SP↗

arXiv2019

Domain adversarial learning for emotion recognition

In practical applications for emotion recognition, users do not always exist in the training corpus. The mismatch between training speakers and testing speakers affects the performance of the trained model. To deal with this problem, we need our model to focus on emotion-related information, while ignoring the difference between speaker identities. In this paper, we look into the use of the domain adversarial neural network (DANN) to extract a common representation between different speakers. The primary task is to predict emotion labels. The secondary task is to learn a common representation where speaker identities can not be distinguished. By using the gradient reversal layer, the gradients coming from the secondary task are used to bring the representations for different speakers closer. To verify the effectiveness of the proposed method, we conduct experiments on the IEMOCAP database. Experimental results demonstrate that the proposed framework shows an absolute improvement of 3.48% over state-of-the-art strategies.

eess.AS↗

arXiv2019

Conversational Emotion Analysis via Attention Mechanisms

Different from the emotion recognition in individual utterances, we propose a multimodal learning framework using relation and dependencies among the utterances for conversational emotion analysis. The attention mechanism is applied to the fusion of the acoustic and lexical features. Then these fusion representations are fed into the self-attention based bi-directional gated recurrent unit (GRU) layer to capture long-term contextual information. To imitate real interaction patterns of different speakers, speaker embeddings are also utilized as additional inputs to distinguish the speaker identities during conversational dialogs. To verify the effectiveness of the proposed method, we conduct experiments on the IEMOCAP database. Experimental results demonstrate that our method shows absolute 2.42% performance improvement over the state-of-the-art strategies.

cs.CL↗

arXiv2019

Unsupervised Representation Learning with Future Observation Prediction for Speech Emotion Recognition

Prior works on speech emotion recognition utilize various unsupervised learning approaches to deal with low-resource samples. However, these methods pay less attention to modeling the long-term dynamic dependency, which is important for speech emotion recognition. To deal with this problem, this paper combines the unsupervised representation learning strategy -- Future Observation Prediction (FOP), with transfer learning approaches (such as Fine-tuning and Hypercolumns). To verify the effectiveness of the proposed method, we conduct experiments on the IEMOCAP database. Experimental results demonstrate that our method is superior to currently advanced unsupervised learning strategies.

eess.AS↗

arXiv2019

Normal Estimation for 3D Point Clouds via Local Plane Constraint and Multi-scale Selection

In this paper, we propose a normal estimation method for unstructured 3D point clouds. In this method, a feature constraint mechanism called Local Plane Features Constraint (LPFC) is used and then a multi-scale selection strategy is introduced. The LPEC can be used in a single-scale point network architecture for a more stable normal estimation of the unstructured 3D point clouds. In particular, it can partly overcome the influence of noise on a large sampling scale compared to the other methods which only use regression loss for normal estimation. For more details, a subnetwork is built after point-wise features extracted layers of the network and it gives more constraints to each point of the local patch via a binary classifier in the end. Then we use multi-task optimization to train the normal estimation and local plane classification tasks simultaneously.Also, to integrate the advantages of multi-scale results, a scale selection strategy is adopted, which is a data-driven approach for selecting the optimal scale around each point and encourages subnetwork specialization. Specifically, we employed a subnetwork called Scale Estimation Network to extract scale weight information from multi-scale features. More analysis is given about the relations between noise levels, local boundary, and scales in the experiment. These relationships can be a better guide to choosing particular scales for a particular model. Besides, the experimental result shows that our network can distinguish the points on the fitting plane accurately and this can be used to guide the normal estimation and our multi-scale method can improve the results well. Compared to some state-of-the-art surface normal estimators, our method is robust to noise and can achieve competitive results.

cs.GR↗

arXiv2019

Semidiscrete quantum droplets and vortices

We consider a binary bosonic condensate with weak mean-field (MF) residual repulsion, loaded in an array of nearly one-dimensional traps coupled by transverse hopping. With the MF force balanced by the effectively one-dimensional attraction, induced in each trap by the Lee-Hung-Yang correction (produced by quantum fluctuations around the MF state), stable onsite-centered and intersite-centered semi-discrete quantum droplets (QDs) emerge in the array, as fundamental ones and self-trapped vortices, with winding numbers, at least, up to 5, in both tightly-bound and quasi-continuum forms. The application of a relatively strong trapping potential leads to squeezing transitions, which increase the number of sites in fundamental QDs, and eventually replace vortex modes by fundamental or dipole ones. The results provide the first realization of stable semi-discrete vortex QDs, including ones with multiple vorticity.

nlin.PS↗

arXiv2019

Expediting hydrogen evolution through topological surface states on Bi2Te3

Recently, the development of efficient and non-noble metal electrocatalysts with excellent durability for the hydrogen evolution reaction (HER) has attracted increasing attention. The exotic and robust metallic surface states of topological insulators (TIs) are theoretically predicted to enhance surface catalytic activity of overlaying catalysts, but no experimental evidence for TIs directly used as electrocatalysts has ever been reported. In this work, we fabricated the TI thin films of Bi2Te3 with different thicknesses using the molecular beam epitaxy method, and found that these thin films exhibit high electrocatalytic activity in HER. The 48 nm Bi2Te3 thin film has the best performance, which is attributed to its largest active area arising from the spiral growth mode of triangular domains as revealed by atomic force microscopy imaging. Importantly, our theoretical calculations reveal that while pure Bi2Te3 is not a good electrocatalyst, the Bi2Te3 thin films with partially oxidized surfaces or Te vacancies have high HER activity. The existence of the corresponding surface oxides on the Bi2Te3 thin films is supported by our X-ray photoelectron spectroscopy data. Particularly, we demonstrate that the topological surface states play a key role in enhancing the HER performance. Our study offers a new direction to design cost-effective electrocatalysts.

physics.chem-ph↗

arXiv2019

Lattice distortion induced first and second order topological phase transition in rectangular high-T$_{\rm c}$ superconducting monolayer

We theoretically study the lattice distortion induced first and second order topological phase transition in rectangular FeSe$_{x}$Te$_{1-x}$ monolayer. When compressing the lattice constant in one direction, our first principles calculation shows that the FeSe$_x$Te$_{1-x}$ undergoes a band inversion at $Γ$ point in a wide dopping range, say $x\in(0.0,0.7)$, which ensures coexistence of the topological band state and the high-T$_{\rm c}$ superconductivity. This unidirectional pressure also leads to the C$_4$ symmetry breaking which is necessary for the monolayer FeSe$_x$Te$_{1-x}$ to support Majorana corner states in the either presence or absence of time-reversal symmetry. Particularly, we use $k\cdot p$ methods to fit the band structure from the first principles calculation and found that the edge states along the $(100)$ and $(010)$ directions have different Dirac energy due to C$_4$ symmetry breaking. This is essential to obtain Majorana corner states in D class without concerning the details of the superconducting pairing symmetries and Zeeman form, which can potentially bring advantages in the experimental implementation.

cond-mat.mes-hall↗

arXiv2019

Temporal Data Fusion at the Edge

As an enabler technique, data fusion has gained great attention in the context of Internet of things (IoT). In traditional settings, data fusion is done at the cloud server. So the data to be fused should be transferred from the sensor nodes to the cloud server before data fusion. Such an application mode of data fusion inherits disturbing concerns from the cloud computing framework, e.g., privacy-leaking, large latency between data capture and computation, excessive ingress bandwidth consumption. We take into account how to do temporal data fusion at the edge to bypass the above issues. We present a Gaussian process based temporal data fusion (GPTDF) method targeted for the problem of sequential online prediction at the edge. The GPTDF method fits the edge computing framework and thus inherits desirable properties from edge computing, such as privacy-preserving, low latency between data capture and computation, and tiny bandwidth consumption. Through a real-data experiment using archived traffic datasets from the Caltrans Performance Measurement System (PeMS), we demonstrate that the application of GPTDF can provide more timely and accurate real-time predictions at the network edge.

cs.DC↗

arXiv2019

Binary Mergers near a Supermassive Black Hole: Relativistic Effects in Triples

We study the general relativitic (GR) effects induced by a spinning supermassive black hole on the orbital and spin evolution of a merging black hole binary (BHB) in a hierarchical triple system. A sufficiently inclined outer orbit can excite Lidov-Kozai eccentricity oscillations in the BHB and induce its merger. These GR effects generate extra precessions on the BHB orbits and spins, significantly increasing the inclination window for mergers and producing a wide range of spin orientations when the BHB enters LIGO band. This "GR-enhanced" channel may play an important role in BHB mergers.

astro-ph.HE↗

arXiv2019

Enhancement of the upper critical field in the cubic Laves-phase superconductor HfV$_{2}$ by Nb doping

We report the effect of Nb doping on the upper critical field of the cubic Laves-phase superconductor HfV$_{2}$ studied in a series of HfV$_{2-x}$Nb$_{x}$ samples with 0 $\leq$ $x$ $\leq$ 0.3 under pulsed fields up to 30 T. The undoped HfV$_{2}$ undergoes a martensitic structural transition around 110 K, and becomes superconducting below $T_{\rm c}$ = 9.4 K. Upon Nb doping, while the structural transition is suppressed for $x$ $\geq$ 0.1, a maximum in $T_{\rm c}$ of 10.1 K and zero-temperature upper critical field $B_{\rm c2}$(0) of 22.4 T is found at $x$ = 0.2, which is ascribed to an increase of the density of states at the Fermi level. For all samples, the temperature dependence of $B_{\rm c2}$ can be well described by the Werthamer-Helfand-Hohenberg (WHH) theory that takes into account both the spin paramagnetic effect and spin orbit scattering. In addition, a comparison is made between the $B_{\rm c2}$ behavior of HfV$_{2-x}$Nb$_{x}$ and those of Nb-Ti and Nb$_{3}$Sn.

cond-mat.supr-con↗

arXiv2019

Memory-Based Neighbourhood Embedding for Visual Recognition

Learning discriminative image feature embeddings is of great importance to visual recognition. To achieve better feature embeddings, most current methods focus on designing different network structures or loss functions, and the estimated feature embeddings are usually only related to the input images. In this paper, we propose Memory-based Neighbourhood Embedding (MNE) to enhance a general CNN feature by considering its neighbourhood. The method aims to solve two critical problems, i.e., how to acquire more relevant neighbours in the network training and how to aggregate the neighbourhood information for a more discriminative embedding. We first augment an episodic memory module into the network, which can provide more relevant neighbours for both training and testing. Then the neighbours are organized in a tree graph with the target instance as the root node. The neighbourhood information is gradually aggregated to the root node in a bottom-up manner, and aggregation weights are supervised by the class relationships between the nodes. We apply MNE on image search and few shot learning tasks. Extensive ablation studies demonstrate the effectiveness of each component, and our method significantly outperforms the state-of-the-art approaches.

cs.CV↗

arXiv2019

Black Hole and Neutron Star Binary Mergers in Triple Systems: II. Merger Eccentricity and Spin-Orbit Misalignment

We study the dynamical signatures of black hole (BH) and neutron star (NS) binary mergers via Lidov-Kozai oscillations induced by tertiary companions in hierarchical triple systems. For each type of binaries (BH-BH and BH-NS), we explore a wide range of binary/triple parameters that lead to binary mergers, and determine the distributions of eccentricity ($e_\mathrm{m}$) and spin-orbit misalignment angle ($θ_\rm {sl}^\rm f$) when the binary enters the LIGO/VIRGO band. We use the double-averaged (over both orbits) and single-averaged (over the inner orbit) secular equations, as well as N-body integration, to evolve systems with different hierarchy levels, including the leading-order post-Newtonian effect, spin-orbit coupling and gravitational radiation. We find that for merging BH-BH binaries with comparable masses, about $7\%$ have $e_\mathrm{m}>0.1$ and $0.7\%$ have $e_{\rm m}>0.9$. The majority of the mergers have significant eccentricities in the LISA band. The BH spin evolution and $θ_\rm {sl}^\rm f$ are correlated with the orbital evolution and $e_{\rm m}$. Mergers with $e_{\rm m} \lesssim 10^{-3}$ have a distribution of $θ_\rm {sl}^\rm f$ that peaks around $90^\circ$ (and thus favoring a projected binary spin parameter $χ_{\rm eff}\sim 0$), while mergers with larger $e_{\rm m}$ have a more isotropic spin-orbit misalignments. For typical BH-NS binaries, strong octuple effects lead to more mergers with non-negligible $e_\mathrm{m}$ (with $\sim 18\%$ having $e_\mathrm{m}>0.1$ and $2.5\%$ having $e_\mathrm{m}>0.9$), and the final BH spin axis tends to be randomly orientated. Measurements or constraints on eccentric mergers and $θ_\rm {sl}^\rm f$ from LIGO/VIRGO and LISA would provide useful diagnostics on the dynamical formation of merging BH or NS binaries in triples.

astro-ph.HE↗

arXiv2019

Deep Metric Transfer for Label Propagation with Limited Annotated Data

We study object recognition under the constraint that each object class is only represented by very few observations. Semi-supervised learning, transfer learning, and few-shot recognition all concern with achieving fast generalization with few labeled data. In this paper, we propose a generic framework that utilizes unlabeled data to aid generalization for all three tasks. Our approach is to create much more training data through label propagation from the few labeled examples to a vast collection of unannotated images. The main contribution of the paper is that we show such a label propagation scheme can be highly effective when the similarity metric used for propagation is transferred from other related domains. We test various combinations of supervised and unsupervised metric learning methods with various label propagation algorithms. We find that our framework is very generic without being sensitive to any specific techniques. By taking advantage of unlabeled data in this way, we achieve significant improvements on all three tasks.

cs.CV↗

arXiv2019

Discriminative Learning for Monaural Speech Separation Using Deep Embedding Features

Deep clustering (DC) and utterance-level permutation invariant training (uPIT) have been demonstrated promising for speaker-independent speech separation. DC is usually formulated as two-step processes: embedding learning and embedding clustering, which results in complex separation pipelines and a huge obstacle in directly optimizing the actual separation objectives. As for uPIT, it only minimizes the chosen permutation with the lowest mean square error, doesn't discriminate it with other permutations. In this paper, we propose a discriminative learning method for speaker-independent speech separation using deep embedding features. Firstly, a DC network is trained to extract deep embedding features, which contain each source's information and have an advantage in discriminating each target speakers. Then these features are used as the input for uPIT to directly separate the different sources. Finally, uPIT and DC are jointly trained, which directly optimizes the actual separation objectives. Moreover, in order to maximize the distance of each permutation, the discriminative learning is applied to fine tuning the whole model. Our experiments are conducted on WSJ0-2mix dataset. Experimental results show that the proposed models achieve better performances than DC and uPIT for speaker-independent speech separation.

cs.SD↗

arXiv2019

Prospects of strongly lensed repeating fast radio bursts: complementary constraints on dark energy evolution

Fast radio bursts (FRBs) are highly dispersed and probably extragalactic radio flashes with millisecond-duration. Recently, the Canadian Hydrogen Intensity Mapping Experiment (using the CHIME/FRB instrument) has reported detections of 13 FRBs during a pre-commissioning phase. It is more exciting that one of the 13 FRBs is a second source of repeaters which suggests that CHIME/FRB and other wide-field sensitive radio telescopes will find a substantial population of repeating FRBs. We have proposed strongly lensed repeating FRBs as a precision cosmological probe, e.g. constraining the Hubble constant and model-independently estimating the cosmic curvature. Here, we study complementary constraints on the equation of state of dark energy from strongly lensed FRBs to currently available popular probes. It is found that, in the framework of Chevalier-Polarski-Linder parametrization, adding time delay measurement of 30 strongly lensed FRB systems to cosmic microwave background radiation and type Ia supernovae can improve the dark energy figure of merit by a factor 2. In the precision cosmology era, this improvement is of great significance for studying the nature of dark energy.

astro-ph.CO↗

arXiv2019

Symmetry breaking of a matter-wave soliton in a double-well potential formed by spatially confined spin-orbit coupling

We consider the symmetry breaking of a matter-wave soliton formed by spinor Bose-Einstein condensates (BECs) illuminated by a two-spot laser beam. This laser beam introduces spin-orbit (SO) coupling in the BECs such that the SO coupling produces an effect similar to a linear doublewell potential (DWP). It is well known that symmetry breaking in a DWP is an important effect and has been discussed in many kinds of systems. However, it has not yet been discussed in a DWP formed by SO coupling. The objective of this work is to study the symmetry breaking of spinor BECs trapped by a DWP formed by SO coupling. We find that two kinds of symmetry breaking, displacement symmetry breaking and bimodal symmetry breaking, can be obtained in this model. The influence of the symmetry transition is systematically discussed by controlling the interaction strength of the BECs and the distance between the center of the two spots. Moreover, because SO coupling violates Galilean invariance, the influence of symmetry breaking in the moving system is also addressed in this paper.

physics.optics↗

arXiv2019

Superconducting phase diagram and nontrivial band topology of structurally modulated Sn$_{1-x}$Sb$_{x}$

We report the discovery of superconductivity in binary alloy Sn$_{1-x}$Sb$_{x}$ with $x$ in the range of 0.43 to 0.6, which possesses a modulated rhombohedral structure due to the incommensurate ordering of Sn and Sb layers along the $c$-axis. The specific heat measurements indicate a weakly coupled, fully gapped superconducting state in this homogeneity range with a maximum bulk $T_{\rm c}$ of 1.58 K at $x$ = 0.46, though the electronic specific heat and Hall coefficients remain nearly $x$-independent. The nonmonotonic dependence of the bulk $T_{\rm c}$ is discussed in relation to the effects of Sb-layer intercalation between the [Sn$_{4}$Sb$_{3}$] seven-layer lamellae that are the essential building block for superconductivity. On the other hand, a zero-resistivity transition is found to take place well above the bulk superconducting transition, and the corresponding $T_{\rm c}$ increases monotonically with $x$ from 2.06 K to 3.29 K. This contrast, together with the uniform elements distribution revealed by energy dispersive x-ray mapping, implies that the resistive transition is due to the strain effect at the grain boundary rather than the compositional inhomogeneity. The first-principles calculations on the representative composition Sn$_{4}$Sb$_{3}$ ($x$ = 0.43) indicate that it is topologically nontrivial similar to Sb, but with different Z$_{2}$ invariants (0;111). Our results not only identify a new superconducting region in the Sn-Sb phase diagram, but also provide a viable platform to study the interplay between structural modulation, nontrivial band topology and superconductivity.

cond-mat.supr-con↗