arXiv subjects

Xin Wang

Publications and source records attributed to Xin Wang.

At least 1,639 records · Page 91Linked to original sources

arXiv2018

Conditional Graph Neural Processes: A Functional Autoencoder Approach

We introduce a novel encoder-decoder architecture to embed functional processes into latent vector spaces. This embedding can then be decoded to sample the encoded functions over any arbitrary domain. This autoencoder generalizes the recently introduced Conditional Neural Process (CNP) model of random processes. Our architecture employs the latest advances in graph neural networks to process irregularly sampled functions. Thus, we refer to our model as Conditional Graph Neural Process (CGNP). Graph neural networks can effectively exploit `local' structures of the metric spaces over which the functions/processes are defined. The contributions of this paper are twofold: (i) a novel graph-based encoder-decoder architecture for functional and process embeddings, and (ii) a demonstration of the importance of using the structure of metric spaces for this type of representations.

cs.LG↗

arXiv2018

Audiovisual speaker conversion: jointly and simultaneously transforming facial expression and acoustic characteristics

An audiovisual speaker conversion method is presented for simultaneously transforming the facial expressions and voice of a source speaker into those of a target speaker. Transforming the facial and acoustic features together makes it possible for the converted voice and facial expressions to be highly correlated and for the generated target speaker to appear and sound natural. It uses three neural networks: a conversion network that fuses and transforms the facial and acoustic features, a waveform generation network that produces the waveform from both the converted facial and acoustic features, and an image reconstruction network that outputs an RGB facial image also based on both the converted features. The results of experiments using an emotional audiovisual database showed that the proposed method achieved significantly higher naturalness compared with one that separately transformed acoustic and facial features.

eess.AS↗

arXiv2018

Guided Feature Selection for Deep Visual Odometry

We present a novel end-to-end visual odometry architecture with guided feature selection based on deep convolutional recurrent neural networks. Different from current monocular visual odometry methods, our approach is established on the intuition that features contribute discriminately to different motion patterns. Specifically, we propose a dual-branch recurrent network to learn the rotation and translation separately by leveraging current Convolutional Neural Network (CNN) for feature representation and Recurrent Neural Network (RNN) for image sequence reasoning. To enhance the ability of feature selection, we further introduce an effective context-aware guidance mechanism to force each branch to distill related information for specific motion pattern explicitly. Experiments demonstrate that on the prevalent KITTI and ICL_NUIM benchmarks, our method outperforms current state-of-the-art model- and learning-based methods for both decoupled and joint camera pose recovery.

cs.CV↗

arXiv2018

Learning to Compose Topic-Aware Mixture of Experts for Zero-Shot Video Captioning

Although promising results have been achieved in video captioning, existing models are limited to the fixed inventory of activities in the training corpus, and do not generalize to open vocabulary scenarios. Here we introduce a novel task, zero-shot video captioning, that aims at describing out-of-domain videos of unseen activities. Videos of different activities usually require different captioning strategies in many aspects, i.e. word selection, semantic construction, and style expression etc, which poses a great challenge to depict novel activities without paired training data. But meanwhile, similar activities share some of those aspects in common. Therefore, We propose a principled Topic-Aware Mixture of Experts (TAMoE) model for zero-shot video captioning, which learns to compose different experts based on different topic embeddings, implicitly transferring the knowledge learned from seen activities to unseen ones. Besides, we leverage external topic-related text corpus to construct the topic embedding for each activity, which embodies the most relevant semantic vectors within the topic. Empirical results not only validate the effectiveness of our method in utilizing semantic knowledge for video captioning, but also show its strong generalization ability when describing novel activities.

cs.CL↗

arXiv2018

Convergence analysis of the block Gibbs sampler for Bayesian probit linear mixed models with improper priors

In this article, we consider Markov chain Monte Carlo(MCMC) algorithms for exploring the intractable posterior density associated with Bayesian probit linear mixed models under improper priors on the regression coefficients and variance components. In particular, we construct the two-block Gibbs sampler using the data augmentation (DA) techniques. Furthermore, we prove geometric ergodicity of the Gibbs sampler, which is the foundation for building central limit theorems for MCMC based estimators and subsequent inferences. The conditions for geometric convergence are similar to those guaranteeing posterior propriety. We also provide conditions for posterior propriety when the design matrices take commonly observed forms. In general, the Haar parameter expansion for DA (PX- DA) algorithm is an improvement of the DA algorithm and it has been shown that it is theoretically at least as good as the DA algorithm. Here we construct a Haar PX-DA algorithm, which has essentially the same computational cost as the two-block Gibbs sampler.

math.ST↗

arXiv2018

Blowup Equations for Refined Topological Strings

Göttsche-Nakajima-Yoshioka K-theoretic blowup equations characterize the Nekrasov partition function of five dimensional $\mathcal{N}=1$ supersymmetric gauge theories compactified on a circle, which via geometric engineering correspond to the refined topological string theory on $SU(N)$ geometries. In this paper, we study the K-theoretic blowup equations for general local Calabi-Yau threefolds. We find that both vanishing and unity blowup equations exist for the partition function of refined topological string, and the crucial ingredients are the $\bf r$ fields introduced in our previous paper. These blowup equations are in fact the functional equations for the partition function and each of them results in infinite identities among the refined free energies. Evidences show that they can be used to determine the full refined BPS invariants of local Calabi-Yau threefolds. This serves an independent and sometimes more powerful way to compute the partition function other than the refined topological vertex in the A-model and the refined holomorphic anomaly equations in the B-model. We study the modular properties of the blowup equations and provide a procedure to determine all the vanishing and unity $\bf r$ fields from the polynomial part of refined topological string at large radius point. We also find that certain form of blowup equations exist at generic loci of the moduli space.

hep-th↗

arXiv2018

Blowup Equations for 6d SCFTs. I

We propose novel functional equations for the BPS partition functions of 6d (1,0) SCFTs, which can be regarded as an elliptic version of Gottsche-Nakajima-Yoshioka's K-theoretic blowup equations. From the viewpoint of geometric engineering, these are the generalized blowup equations for refined topological strings on certain local elliptic Calabi-Yau threefolds. We derive recursion formulas for elliptic genera of self-dual strings on the tensor branch from these functional equations and in this way obtain a universal approach for determining refined BPS invariants. As examples, we study in detail the minimal 6d SCFTs with SU(3) and SO(8) gauge symmetry. In companion papers, we will study the elliptic blowup equations for all other non-Higgsable clusters.

hep-th↗

arXiv2018

A Deep Dive into Blockchain Selfish Mining

This paper studies a fundamental problem regarding the security of blockchain on how the existence of multiple misbehaving pools influences the profitability of selfish mining. Each selfish miner maintains a private chain and makes it public opportunistically for the purpose of acquiring more rewards incommensurate to his Hashrate. We establish a novel Markov chain model to characterize all the state transitions of public and private chains. The minimum requirement of Hashrate together with the minimum delay of being profitable is derived in close-form. The former reduces to 21.48% with the symmetric selfish miners, while their competition with asymmetric Hashrates puts forward a higher requirement of the profitable threshold. The profitable delay increases with the decrease of the Hashrate of selfish miners, making the mining pools more cautious on performing selfish mining.

cs.CR↗

arXiv2018

STFT spectral loss for training a neural speech waveform model

This paper proposes a new loss using short-time Fourier transform (STFT) spectra for the aim of training a high-performance neural speech waveform model that predicts raw continuous speech waveform samples directly. Not only amplitude spectra but also phase spectra obtained from generated speech waveforms are used to calculate the proposed loss. We also mathematically show that training of the waveform model on the basis of the proposed loss can be interpreted as maximum likelihood training that assumes the amplitude and phase spectra of generated speech waveforms following Gaussian and von Mises distributions, respectively. Furthermore, this paper presents a simple network architecture as the speech waveform model, which is composed of uni-directional long short-term memories (LSTMs) and an auto-regressive structure. Experimental results showed that the proposed neural model synthesized high-quality speech waveforms.

eess.AS↗

arXiv2018

Learning and Management for Internet-of-Things: Accounting for Adaptivity and Scalability

Internet-of-Things (IoT) envisions an intelligent infrastructure of networked smart devices offering task-specific monitoring and control services. The unique features of IoT include extreme heterogeneity, massive number of devices, and unpredictable dynamics partially due to human interaction. These call for foundational innovations in network design and management. Ideally, it should allow efficient adaptation to changing environments, and low-cost implementation scalable to massive number of devices, subject to stringent latency constraints. To this end, the overarching goal of this paper is to outline a unified framework for online learning and management policies in IoT through joint advances in communication, networking, learning, and optimization. From the network architecture vantage point, the unified framework leverages a promising fog architecture that enables smart devices to have proximity access to cloud functionalities at the network edge, along the cloud-to-things continuum. From the algorithmic perspective, key innovations target online approaches adaptive to different degrees of nonstationarity in IoT dynamics, and their scalable model-free implementation under limited feedback that motivates blind or bandit approaches. The proposed framework aspires to offer a stepping stone that leads to systematic designs and analysis of task-specific learning and management schemes for IoT, along with a host of new research directions to build on.

eess.SY↗

arXiv2018

Using and reusing coherence to realize quantum processes

Coherent superposition is a key feature of quantum mechanics that underlies the advantage of quantum technologies over their classical counterparts. Recently, coherence has been recast as a resource theory in an attempt to identify and quantify it in an operationally well-defined manner. Here we study how the coherence present in a state can be used to implement a quantum channel via incoherent operations and, in turn, to assess its degree of coherence. We introduce the robustness of coherence of a quantum channel---which reduces to the homonymous measure for states when computed on constant-output channels---and prove that: i) it quantifies the minimal rank of a maximally coherent state required to implement the channel; ii) its logarithm quantifies the amortized cost of implementing the channel provided some coherence is recovered at the output; iii) its logarithm also quantifies the zero-error asymptotic cost of implementation of many independent copies of a channel. We also consider the generalized problem of imperfect implementation with arbitrary resource states. Using the robustness of coherence, we find that in general a quantum channel can be implemented without employing a maximally coherent resource state. In fact, we prove that \textit{every} pure coherent state in dimension larger than $2$, however weakly so, turns out to be a valuable resource to implement \textit{some} coherent unitary channel. We illustrate our findings for the case of single-qubit unitary channels.

quant-ph↗

arXiv2018

Nonlinear dynamics of Shear Alfvén fluctuations in Divertor Tokamak Test facility plasmas

Following the analysis on linear spectra of shear Alfvén fluctuations excited by energetic particles (EPs) in the Divertor Tokamak Test (DTT) facility plasmas [T. Wang et al., Phys. Plasmas 25, 062509 (2018)], in this work, nonlinear dynamics of the corresponding mode saturation and the fluctuation induced EP transport is studied by hybrid magnetohydrodynamic-gyrokinetic simulations. For the reversed shear Alfvén eigenmode driven by magnetically trapped EP precession resonance in the central core region of DTT plasmas, the saturation is mainly due to radial decoupling of resonant trapped EPs. Consistent with the wave-EP resonance structure, EP transport occurs in a similar scale to the mode width. On the other hand, passing EP transport is analyzed in detail for toroidal Alfvén eigenmode in the outer core region, with mode drive from both passing and trapped EPs. It is shown that passing EPs experience only weak redistributions in the weakly unstable case; and the transport extends to meso-scale diffusion in the strongly unstable case, due to orbit stochasticity induced by resonance overlap. Here, weakly/strongly unstable regime is determined by Chirikov condition for resonance overlap. This work then further illuminates rich and diverse nonlinear EP dynamics related to burning plasma studies, and the capability of DTT to address these key physics.

physics.plasm-ph↗

arXiv2018

Mass and Light of Abell 370: A Strong and Weak Lensing Analysis

We present a new gravitational lens model of the Hubble Frontier Fields cluster Abell 370 ($z = 0.375$) using imaging and spectroscopy from Hubble Space Telescope and ground-based spectroscopy. We combine constraints from a catalog of 1344 weakly lensed galaxies and 39 multiply-imaged sources comprised of 114 multiple images, including a system of multiply-imaged candidates at $z=7.93 \pm 0.02$, to obtain a best-fit mass distribution using the cluster lens modeling code Strong and Weak Lensing United. As the only analysis of A370 using strong and weak lensing constraints from Hubble Frontier Fields data, our method provides an independent check on assumptions in other methods on the mass distribution. Convergence, shear, and magnification maps are made publicly available through the HFF website. We find that the model we produce is similar to models produced by other groups, with some exceptions due to the differences in lensing code methodology. In an effort to study how our total projected mass distribution traces light, we measure the stellar mass density distribution using Spitzer/Infrared Array Camera imaging. Comparing our total mass density to our stellar mass density in a radius of 0.3 Mpc, we find a mean projected stellar to total mass ratio of $\langle f* \rangle = 0.011 \pm 0.003$ (stat.) using the diet Salpeter initial mass function. This value is in general agreement with independent measurements of $\langle f* \rangle$ in clusters of similar total mass and redshift.

astro-ph.GA↗

arXiv2018

Tunable charge qubit based on barrier-controlled triple quantum dots

We present a theoretical proposal of a tunable charge qubit, hosted in triple quantum dots. The manipulation is solely performed by changing the heights of the two potential barriers between the three dots, while the energy of all three dots are fixed. We have found that when the relative height of the two barriers are changed, the direction of the axis of rotation in performing single-qubit gates can be varied. On the other hand, the corresponding rotation speed can be tuned by raising or lowering the two barriers at the same time. Our proposal therefore allows for tunability of both the rotation axis and rotating speed for a charge qubit via all-electrical control, which may facilitate realization of quantum algorithms in these devices.

cond-mat.mes-hall↗

arXiv2018

Semidefinite programming converse bounds for quantum communication

We derive several efficiently computable converse bounds for quantum communication over quantum channels in both the one-shot and asymptotic regime. First, we derive one-shot semidefinite programming (SDP) converse bounds on the amount of quantum information that can be transmitted over a single use of a quantum channel, which improve the previous bound from [Tomamichel/Berta/Renes, Nat. Commun. 7, 2016]. As applications, we study quantum communication over depolarizing channels and amplitude damping channels with finite resources. Second, we find an SDP strong converse bound for the quantum capacity of an arbitrary quantum channel, which means the fidelity of any sequence of codes with a rate exceeding this bound will vanish exponentially fast as the number of channel uses increases. Furthermore, we prove that the SDP strong converse bound improves the partial transposition bound introduced by Holevo and Werner. Third, we prove that this SDP strong converse bound is equal to the so-called max-Rains information, which is an analog to the Rains information introduced in [Tomamichel/Wilde/Winter, IEEE Trans. Inf. Theory 63:715, 2017]. Our SDP strong converse bound is weaker than the Rains information, but it is efficiently computable for general quantum channels.

quant-ph↗

arXiv2018

Deep Encoder-Decoder Models for Unsupervised Learning of Controllable Speech Synthesis

Generating versatile and appropriate synthetic speech requires control over the output expression separate from the spoken text. Important non-textual speech variation is seldom annotated, in which case output control must be learned in an unsupervised fashion. In this paper, we perform an in-depth study of methods for unsupervised learning of control in statistical speech synthesis. For example, we show that popular unsupervised training heuristics can be interpreted as variational inference in certain autoencoder models. We additionally connect these models to VQ-VAEs, another, recently-proposed class of deep variational autoencoders, which we show can be derived from a very similar mathematical argument. The implications of these new probabilistic interpretations are discussed. We illustrate the utility of the various approaches with an application to acoustic modelling for emotional speech synthesis, where the unsupervised methods for learning expression control (without access to emotional labels) are found to give results that in many aspects match or surpass the previous best supervised approach.

eess.AS↗

arXiv2018

Spontaneous Edge Current in Higher Chirality Superconductors

The effects of finite temperature, Meissner screening and surface roughness on the spontaneous edge current for higher chirality quasi-two dimensional superconductors are studied in the continuum limit using the quasiclassical Eilenberger equations. We find that the total spontaneous current is non-zero at finite temperature $T$ and maximized near $T=T_c/2$, where $T_c$ is the transition temperature, although it vanishes at $T=0$. In the presence of surface roughness, we observe a surface current inversion in the chiral $d$-wave case that can be understood in terms of a disorder induced $s$-wave pairing component in the rough surface regime. This conclusion is supported by a Ginzburg-Landau analysis. However, this current inversion is non-universal beyond the continuum limit as demonstrated by self-consistent lattice Bogoliubov-de Gennes calculations.

cond-mat.supr-con↗

arXiv2018

Gaussian quantum resource theories

We develop a general framework to assess capabilities and limitations of the Gaussian toolbox in continuous variable quantum information theory. Our framework allows us to characterize the structure and properties of quantum resource theories specialized to Gaussian states and Gaussian operations, establishing rigorous methods for their description and yielding a unified approach to their quantification. We show in particular that, under a few intuitive and physically motivated assumptions on the set of free states, no Gaussian quantum resource can be distilled with free Gaussian operations, even when an unlimited supply of the resource state is available. This places fundamental constraints on state manipulations in all such Gaussian resource theories. We discuss in particular the applications to quantum entanglement, where we extend previously known results by showing that Gaussian entanglement cannot be distilled even with Gaussian operations preserving the positivity of the partial transpose, as well as to other Gaussian resources such as steering and optical nonclassicality. A comprehensive semidefinite programming representation of all these resources is explicitly provided.

quant-ph↗