arXiv subjects

Can Xu

Publications and source records attributed to Can Xu.

123 records · Page 7Linked to original sources

arXiv2019

Latent Class Model with Application to Speaker Diarization

In this paper, we apply a latent class model (LCM) to the task of speaker diarization. LCM is similar to Patrick Kenny's variational Bayes (VB) method in that it uses soft information and avoids premature hard decisions in its iterations. In contrast to the VB method, which is based on a generative model, LCM provides a framework allowing both generative and discriminative models. The discriminative property is realized through the use of i-vector (Ivec), probabilistic linear discriminative analysis (PLDA), and a support vector machine (SVM) in this work. Systems denoted as LCM-Ivec-PLDA, LCM-Ivec-SVM, and LCM-Ivec-Hybrid are introduced. In addition, three further improvements are applied to enhance its performance. 1) Adding neighbor windows to extract more speaker information for each short segment. 2) Using a hidden Markov model to avoid frequent speaker change points. 3) Using an agglomerative hierarchical cluster to do initialization and present hard and soft priors, in order to overcome the problem of initial sensitivity. Experiments on the National Institute of Standards and Technology Rich Transcription 2009 speaker diarization database, under the condition of a single distant microphone, show that the diarization error rate (DER) of the proposed methods has substantial relative improvements compared with mainstream systems. Compared to the VB method, the relative improvements of LCM-Ivec-PLDA, LCM-Ivec-SVM, and LCM-Ivec-Hybrid systems are 23.5%, 27.1%, and 43.0%, respectively. Experiments on our collected database, CALLHOME97, CALLHOME00 and SRE08 short2-summed trial conditions also show that the proposed LCM-Ivec-Hybrid system has the best overall performance.

eess.AS↗

arXiv2019

Towards Neural Mixture Recommender for Long Range Dependent User Sequences

Understanding temporal dynamics has proved to be highly valuable for accurate recommendation. Sequential recommenders have been successful in modeling the dynamics of users and items over time. However, while different model architectures excel at capturing various temporal ranges or dynamics, distinct application contexts require adapting to diverse behaviors. In this paper we examine how to build a model that can make use of different temporal ranges and dynamics depending on the request context. We begin with the analysis of an anonymized Youtube dataset comprising millions of user sequences. We quantify the degree of long-range dependence in these sequences and demonstrate that both short-term and long-term dependent behavioral patterns co-exist. We then propose a neural Multi-temporal-range Mixture Model (M3) as a tailored solution to deal with both short-term and long-term dependencies. Our approach employs a mixture of models, each with a different temporal range. These models are combined by a learned gating mechanism capable of exerting different model combinations given different contextual information. In empirical evaluations on a public dataset and our own anonymized YouTube dataset, M3 consistently outperforms state-of-the-art sequential recommendation methods.

cs.LG↗

arXiv2018

Multiobjective Optimization Training of PLDA for Speaker Verification

Most current state-of-the-art text-independent speaker verification systems take probabilistic linear discriminant analysis (PLDA) as their backend classifiers. The parameters of PLDA are often estimated by maximizing the objective function, which focuses on increasing the value of log-likelihood function, but ignoring the distinction between speakers. In order to better distinguish speakers, we propose a multi-objective optimization training for PLDA. Experiment results show that the proposed method has more than 10% relative performance improvement in both EER and MinDCF on the NIST SRE14 i-vector challenge dataset, and about 20% relative performance improvement in EER on the MCE18 dataset.

cs.SD↗

arXiv2018

Towards Explainable and Controllable Open Domain Dialogue Generation with Dialogue Acts

We study open domain dialogue generation with dialogue acts designed to explain how people engage in social chat. To imitate human behavior, we propose managing the flow of human-machine interactions with the dialogue acts as policies. The policies and response generation are jointly learned from human-human conversations, and the former is further optimized with a reinforcement learning approach. With the dialogue acts, we achieve significant improvement over state-of-the-art methods on response quality for given contexts and dialogue length in both machine-machine simulation and human-machine conversation.

cs.CL↗

arXiv2017

Neural Response Generation with Dynamic Vocabularies

We study response generation for open domain conversation in chatbots. Existing methods assume that words in responses are generated from an identical vocabulary regardless of their inputs, which not only makes them vulnerable to generic patterns and irrelevant noise, but also causes a high cost in decoding. We propose a dynamic vocabulary sequence-to-sequence (DVS2S) model which allows each input to possess their own vocabulary in decoding. In training, vocabulary construction and response generation are jointly learned by maximizing a lower bound of the true objective with a Monte Carlo sampling method. In inference, the model dynamically allocates a small vocabulary for an input with the word prediction model, and conducts decoding only with the small vocabulary. Because of the dynamic vocabulary mechanism, DVS2S eludes many generic patterns and irrelevant words in generation, and enjoys efficient decoding at the same time. Experimental results on both automatic metrics and human annotations show that DVS2S can significantly outperform state-of-the-art methods in terms of response quality, but only requires 60% decoding time compared to the most efficient baseline.

cs.CL↗

arXiv2017

A Sequential Matching Framework for Multi-turn Response Selection in Retrieval-based Chatbots

We study the problem of response selection for multi-turn conversation in retrieval-based chatbots. The task requires matching a response candidate with a conversation context, whose challenges include how to recognize important parts of the context, and how to model the relationships among utterances in the context. Existing matching methods may lose important information in contexts as we can interpret them with a unified framework in which contexts are transformed to fixed-length vectors without any interaction with responses before matching. The analysis motivates us to propose a new matching framework that can sufficiently carry the important information in contexts to matching and model the relationships among utterances at the same time. The new framework, which we call a sequential matching framework (SMF), lets each utterance in a context interacts with a response candidate at the first step and transforms the pair to a matching vector. The matching vectors are then accumulated following the order of the utterances in the context with a recurrent neural network (RNN) which models the relationships among the utterances. The context-response matching is finally calculated with the hidden states of the RNN. Under SMF, we propose a sequential convolutional network and sequential attention network and conduct experiments on two public data sets to test their performance. Experimental results show that both models can significantly outperform the state-of-the-art matching methods. We also show that the models are interpretable with visualizations that provide us insights on how they capture and leverage the important information in contexts for matching.

cs.CL↗

arXiv2017

Order parameter analysis of synchronization transitions on star networks

Collective behaviors of populations of coupled oscillators have attracted much attention in recent years. In this paper, an order parameter approach is proposed to study the low-dimensional dynam- ical mechanism of collective synchronizations by adopting the star-topology of coupled oscillators as a prototype system. The order parameter equation of star-linked phase oscillators can be obtained in terms of the Watanabe-Strogatz transformation, Ott-Antonsen ansatz, and the ensemble order parameter approach. Different solutions of the order parameter equation correspond to diverse col- lective states, and different bifurcations reveal various transitions among these collective states. The properties of various transitions are revealed in the star-network model by using tools of nonlinear dynamics such as time reversibility analysis and linear stability analysis.

nlin.AO↗

arXiv2016

Disorder-driven topological phase transition in Bi2Se3 films

Topological insulators (TI) are a phase of matter that host unusual metallic states on their surfaces. Unlike the states that exist on the surface of conventional materials, these so-called topological surfaces states (TSS) are protected against disorder-related localization effects by time reversal symmetry through strong spin-orbit coupling. By combining transport measurements, angle-resolved photo-emission spectroscopy and scanning tunneling microscopy, we show that there exists a critical level of disorder beyond which the TI Bi2Se3 loses its ability to protect the metallic TSS and transitions to a fully insulating state. The absence of the metallic surface channels dictates that there is a change in topological character, implying that disorder can lead to a topological phase transition even without breaking the time reversal symmetry. This observation challenges the conventional notion of topologically-protected surface states, and will provoke new studies as to the fundamental nature of topological phase of matter in the presence of disorder.

cond-mat.mtrl-sci↗

arXiv2016

Dynamics of phase oscillators in the Kuramoto model with generalized frequency-weighted coupling

We generalize the Kuramoto model for the synchronization transition of globally coupled phase oscillators to populations by incorporating an additional heterogeneity with the coupling strength, where each oscillator pair interacts with different coupling strength weighted by a genera; function of their natural frequency. The expression for the critical coupling can be straightforwardly extended to a generalized explicit formula analytically, and s self-consistency approach is developed to predict the stationary states in the thermodynamic limit. The landau damping effect is further revealed by means of the linear stability analysis and resonance poles theory above the critical threshold which turns to be far more generic. Furthermore, the dimensionality reduction technique of the Ott-Antonsen is implemented to capture the analytical description of relaxation dynamics of the steady states valid on a globally attracting manifold. Our theoretical analysis and numerical results are consistent with each other, which can help us understand the synchronization transition in general networks with heterogenous couplings.

nlin.AO↗

arXiv2016

Collective dynamics of identical phase oscillators with high-order coupling

In this paper, we propose a framework to investigate the collective dynamics in ensembles of globally coupled phase oscillators when higher-order modes dominate the coupling. The spatiotemporal properties of the attractors in various regions of parameter space are analyzed. Furthermore, a detailed linear stability analysis proves that the stationary symmetric distribution is only neutrally stable in the marginal regime which stems from the generalized time-reversal symmetry. Moreover, the critical parameters of the transition among various regimes are determined analytically by both the Ott-Antonsen method and linear stability analysis, the transient dynamics are further revealed in terms of the characteristic curves method. Finally, for the more general initial condition the symmetric dynamics could be reduced to a rigorous three-dimensional manifold which shows that the neutrally stable chaos could also occur in this model for particular parameter. Our theoretical analysis and numerical results are consistent with each other, which can help us understand the dynamical properties in general system with higher-order harmonics couplings.

nlin.AO↗

arXiv2015

Synchronization of phase oscillators with frequency-weighted coupling

Recently, the first-order synchronization transition has been studied in systems of coupled phase oscillators. In this paper, we propose a framework to investigate the synchronization in the frequency-weighted Kuramoto model with all-to-all couplings. A rigorous mean-field analysis is implemented to predict the possible steady states. Furthermore, a detailed linear stability analysis proves that the incoherent state is only neutrally stable below the synchronization threshold. Nevertheless, interestingly, the amplitude of the order parameter decays exponentially (at least for short time) in this regime, resembling the Landau damping effect in plasma physics. Moreover, the explicit expression for the critical coupling strength is determined by both the mean-field method and linear operator theory. The mechanism of bifurcation for the incoherent state near the critical point is further revealed by the amplitude expansion theory, which shows that the oscillating standing wave state could also occur in this model for certain frequency distributions. Our theoretical analysis and numerical results are consistent with each other, which can help us understand the synchronization transition in general networks with heterogenous couplings.

nlin.AO↗

arXiv2015

Order parameter analysis for low-dimensional behaviors of coupled phase-oscillators

Coupled phase-oscillators are important models related to synchronization. Recently, Ott-Antonsen(OA) ansatz is developed and used to get low-dimensional collective behaviors in coupled oscillator systems. In this paper, we develop a simple and concise approach based on the equations of order parameters, namely, order parameter analysis, with which we point out that the OA ansatz is rooted in the dynamical symmetry of the order parameters. With our approach the scope of the OA ansatz is identified as two conditions, i.e., infinite size of the system and only three nonzero Fourier coefficients of the coupling function. Coinciding with each of the conditions, a distinctive system out of the scope is taken into account and discussed with the order parameter analysis. Two approximation methods are introduced respectively, namely the ensemble approach and the dominating-term assumption.

nlin.AO↗

arXiv2015

Ensemble order parameter equations in star network

The OA ansatz has attracted much attention recently, infinite-dimensional Kuramoto model could collapses to a two-dimensional system of order differential equations with it. In this paper, we propose the ensemble order parameter (EOP) equations to describe the dynamics for networks with a finite size. To verify the effectiveness of this method, we apply it into the star network and star-connected network. In the star network, numerous phase transitions among different synchronous states are observed, three processes of synchronization, one process of de-synchronization and a group of hybrid phase transitions, the processes of those transitions are revealed by the EOP dynamics and other nolinear tools such as time reversibility analysis and linear stability analysis. Also in the star-connected network, the two-step synchronization transition is observed. The process of it is still be revealed by the similar methods in the single star network.

nlin.AO↗

arXiv2014

Explosive or Continuous: Incoherent state determines the route to synchronization

Collective behaviors of coupled oscillators have attracted much attention. In this Letter, we propose an ensemble order parameter(EOP) equation that enables us to grasp the essential low-dimensional dynamical mechanism of the explosive synchronization in heterogeneous networks. Dif- ferent solutions of the EOP equation build correspondences with diverse collective states, and different bifurcations reveal various transitions among these collective states. The structural relationship between the incoherent state and synchronous state leads to different routes of transitions to synchronization, either continuous or discontinuous. The explosive synchronization is determined by the bistable state where the measure of each state and the critical points are obtained analytically by using the EOP equation. Our method and results hold for heterogeneous networks with star graph motifs such as scale-free networks, and hence, provide an effective approach in understanding the routes to synchronization in more general complex networks.

nlin.AO↗

arXiv2014

Visual Sentiment Prediction with Deep Convolutional Neural Networks

Images have become one of the most popular types of media through which users convey their emotions within online social networks. Although vast amount of research is devoted to sentiment analysis of textual data, there has been very limited work that focuses on analyzing sentiment of image data. In this work, we propose a novel visual sentiment prediction framework that performs image understanding with Deep Convolutional Neural Networks (CNN). Specifically, the proposed sentiment prediction framework performs transfer learning from a CNN with millions of parameters, which is pre-trained on large-scale data for object recognition. Experiments conducted on two real-world datasets from Twitter and Tumblr demonstrate the effectiveness of the proposed visual sentiment analysis framework.

cs.CV↗