arXiv subjects

Qi Zhang

Publications and source records attributed to Qi Zhang.

At least 1,045 records · Page 58Linked to original sources

arXiv2020

Rethinking Generalization of Neural Models: A Named Entity Recognition Case Study

While neural network-based models have achieved impressive performance on a large body of NLP tasks, the generalization behavior of different models remains poorly understood: Does this excellent performance imply a perfect generalization model, or are there still some limitations? In this paper, we take the NER task as a testbed to analyze the generalization behavior of existing models from different perspectives and characterize the differences of their generalization abilities through the lens of our proposed measures, which guides us to better design models and training methods. Experiments with in-depth analyses diagnose the bottleneck of existing neural NER models in terms of breakdown performance analysis, annotation errors, dataset bias, and category relationships, which suggest directions for improvement. We have released the datasets: (ReCoNLL, PLONER) for the future research at our project page: http://pfliu.com/InterpretNER/. As a by-product of this paper, we have open-sourced a project that involves a comprehensive summary of recent NER papers and classifies them into different research topics: https://github.com/pfliu-nlp/Named-Entity-Recognition-NER-Papers.

cs.CL↗

arXiv2020

Tensor network approach to phase transitions of a non-Abelian topological phase

The non-abelian topological phase with Fibonacci anyons minimally supports universal quantum computation. In order to investigate the possible phase transitions out of the Fibonacci topological phase, we propose a generic quantum-net wavefunction with two tuning parameters dual with each other, and the norm can be exactly mapped into a partition function of the two-coupled $ϕ^{2}$-state Potts models, where $ϕ=(\sqrt{5}+1)/2$ is the golden ratio. By developing the tensor network representation of this wavefunction on a square lattice, we can accurately calculate the full phase diagram with the numerical methods of tensor networks. More importantly, it is found that the non-abelian Fibonacci topological phase is enclosed by three distinct non-topological phases and their dual phases of a single $ϕ^{2}$-state Potts model: the gapped dilute net phase, critical dense net phase, and spontaneous translation symmetry breaking gapped phase. We also determine the critical properties of the phase transitions among the Fibonacci topological phase and those non-topological phases.

cond-mat.str-el↗

arXiv2020

Branch-and-Price for a Class of Nonconvex Mixed-Integer Nonlinear Programs

This work attempts to combine the strengths of two major technologies that have matured over the last three decades: global mixed-integer nonlinear optimization and branch-and-price. We consider a class of generally nonconvex mixed-integer nonlinear programs (MINLPs) with linear complicating constraints and integer linking variables. If the complicating constraints are removed, the problem becomes easy to solve, e.g. due to decomposable structure. Integrality of the linking variables allows us to apply a discretization approach to derive a Dantzig-Wolfe reformulation and solve the problem to global optimality using branch-and-price. It is a remarkably simple idea; but to our surprise, it has barely found any application in the literature. In this work, we show that many relevant problems directly fall or can be reformulated into this class of MINLPs. We present the branch-and-price algorithm and demonstrate its effectiveness (and sometimes ineffectiveness) in an extensive computational study considering multiple large-scale problems of practical relevance, showing that, in many cases, orders-of-magnitude reductions in solution time can be achieved.

math.OC↗

arXiv2019

Multidimensional Variational Line Spectra Estimation

The fundamental multidimensional line spectral estimation problem is addressed utilizing the Bayesian methods. Motivated by the recently proposed variational line spectral estimation (VALSE) algorithm, multidimensional VALSE (MDVALSE) is developed. MDVALSE inherits the advantages of VALSE such as automatically estimating the model order, noise variance and providing uncertain degrees of frequency estimates. Compared to VALSE, the multidimensional frequencies of a single component is treated as a whole, and the probability density function is projected as independent univariate von Mises distribution to perform tractable inference. Besides, for the initialization, efficient fast Fourier transform (FFT) is adopted to significantly reduce the computation complexity of MDVALSE. Numerical results demonstrate the effectiveness of the MDVALSE, compared to state-of-art methods.

cs.IT↗

arXiv2019

Global Solution and Blow-up of the Stochastic Nonlinear Schrödinger system

We study the stochastic Nonlinear Schrödinger system with multiplicative white noise in energy space $H^1$. Based on deterministic and stochastic Strichartz estimates, we prove the local well-posedness and uniqueness of mild solution. Then we prove the global well-posedness in the mass subcritical case and the defocusing case. For the mass subcritical case, we also investigate the global existence when the $L^2$ norm of initial value is small enough. In addition, we study the blow-up phenomenon and give a sharp criteria.

math.AP↗

arXiv2019

Self-driving scale car trained by Deep reinforcement learning

The self-driving based on deep reinforcement learning, as the most important application of artificial intelligence, has become a popular topic. Most of the current self-driving methods focus on how to directly learn end-to-end self-driving control strategy from the raw sensory data. Essentially, this control strategy can be considered as a mapping between images and driving behavior, which usually faces a problem of low generalization ability. To improve the generalization ability for the driving behavior, the reinforcement learning method requires extrinsic reward from the real environment, which may damage the car. In order to obtain a good generalization ability in safety, a virtual simulation environment that can be constructed different driving scene is designed by Unity. A theoretical model is established and analyzed in the virtual simulation environment, and it is trained by double Deep Q-network. Then, the trained model is migrated to a scale car in real world. This process is also called a sim2real method. The sim2real training method efficiently handle the these two problems. The simulations and experiments are carried out to evaluate the performance and effectiveness of the proposed algorithm. Finally, it is demonstrated that the scale car in real world obtain the capability for autonomous driving.

cs.LG↗

arXiv2019

Matrix Completion from Quantized Samples via Generalized Sparse Bayesian Learning

The recovery of a low rank matrix from a subset of noisy low-precision quantized samples arises in several applications such as collaborative filtering, intelligent recommendation and millimeter wave channel estimation with few bit ADCs. In this paper, a generalized sparse Bayesian learning (Gr-SBL) combining expectation propagation (EP) is proposed to solve the matrix completion (MC), which is termed as MC-Gr-SBL. The MC-Gr-SBL automatically estimates the rank, the factors and their covariance matrices, and the noise variance. In addition, MC-Gr-SBL is proposed to solve the two dimensional line spectral estimation problem by incorporating the MUSIC algorithm. Finally, substantial numerical experiments are conducted to verify the effectiveness of the proposed algorithm.

cs.IT↗

arXiv2019

Constructing Multiple Tasks for Augmentation: Improving Neural Image Classification With K-means Features

Multi-task learning (MTL) has received considerable attention, and numerous deep learning applications benefit from MTL with multiple objectives. However, constructing multiple related tasks is difficult, and sometimes only a single task is available for training in a dataset. To tackle this problem, we explored the idea of using unsupervised clustering to construct a variety of auxiliary tasks from unlabeled data or existing labeled data. We found that some of these newly constructed tasks could exhibit semantic meanings corresponding to certain human-specific attributes, but some were non-ideal. In order to effectively reduce the impact of non-ideal auxiliary tasks on the main task, we further proposed a novel meta-learning-based multi-task learning approach, which trained the shared hidden layers on auxiliary tasks, while the meta-optimization objective was to minimize the loss on the main task, ensuring that the optimizing direction led to an improvement on the main task. Experimental results across five image datasets demonstrated that the proposed method significantly outperformed existing single task learning, semi-supervised learning, and some data augmentation methods, including an improvement of more than 9% on the Omniglot dataset.

cs.CV↗

arXiv2019

Authenticated Key-Value Stores with Hardware Enclaves

Authenticated data storage on an untrusted platform is an important computing paradigm for cloud applications ranging from big-data outsourcing, to cryptocurrency and certificate transparency log. These modern applications increasingly feature update-intensive workloads, whereas existing authenticated data structures (ADSs) designed with in-place updates are inefficient to handle such workloads. In this paper, we address this issue and propose a novel authenticated log-structured merge tree (eLSM) based key-value store by leveraging Intel SGX enclaves. We present a system design that runs the code of eLSM store inside enclave. To circumvent the limited enclave memory (128 MB with the latest Intel CPUs), we propose to place the memory buffer of the eLSM store outside the enclave and protect the buffer using a new authenticated data structure by digesting individual LSM-tree levels. We design protocols to support query authentication in data integrity, completeness (under range queries), and freshness. The proof in our protocol is made small by including only the Merkle proofs at selective levels. We implement eLSM on top of Google LevelDB and Facebook RocksDB with minimal code change and performance interference. We evaluate the performance of eLSM under the YCSB workload benchmark and show a performance advantage of up to 4.5X speedup.

cs.CR↗

arXiv2019

Discrete Argument Representation Learning for Interactive Argument Pair Identification

In this paper, we focus on extracting interactive argument pairs from two posts with opposite stances to a certain topic. Considering opinions are exchanged from different perspectives of the discussing topic, we study the discrete representations for arguments to capture varying aspects in argumentation languages (e.g., the debate focus and the participant behavior). Moreover, we utilize hierarchical structure to model post-wise information incorporating contextual knowledge. Experimental results on the large-scale dataset collected from CMV show that our proposed framework can significantly outperform the competitive baselines. Further analyses reveal why our model yields superior performance and prove the usefulness of our learned representations.

cs.CL↗

arXiv2019

Efficient Communication in Multi-Agent Reinforcement Learning via Variance Based Control

Multi-agent reinforcement learning (MARL) has recently received considerable attention due to its applicability to a wide range of real-world applications. However, achieving efficient communication among agents has always been an overarching problem in MARL. In this work, we propose Variance Based Control (VBC), a simple yet efficient technique to improve communication efficiency in MARL. By limiting the variance of the exchanged messages between agents during the training phase, the noisy component in the messages can be eliminated effectively, while the useful part can be preserved and utilized by the agents for better performance. Our evaluation using a challenging set of StarCraft II benchmarks indicates that our method achieves $2-10\times$ lower in communication overhead than state-of-the-art MARL algorithms, while allowing agents to better collaborate by developing sophisticated strategies.

cs.LG↗

arXiv2019

Terahertz emission from anomalous Hall effect in a single-layer ferromagnet

We report on terahertz emission from a single layer ferromagnet which involves the generation of backflow nonthermal charge current from the ferromagnet/dielectric interface by femtosecond laser excitation and subsequent conversion of the charge current to a transverse transient charge current via the anomalous Hall effect, thereby generating the THz radiation. The THz emission can be either enhanced or suppressed, or even the polarity can be reversed, by introducing a magnetization gradient in the thickness direction of the ferromagnet. Unlike spintronic THz emitters reported previously, it does not require additional non-magnetic layer or Rashba interface.

physics.app-ph↗

arXiv2019

Demystifying Learning Rate Policies for High Accuracy Training of Deep Neural Networks

Learning Rate (LR) is an important hyper-parameter to tune for effective training of deep neural networks (DNNs). Even for the baseline of a constant learning rate, it is non-trivial to choose a good constant value for training a DNN. Dynamic learning rates involve multi-step tuning of LR values at various stages of the training process and offer high accuracy and fast convergence. However, they are much harder to tune. In this paper, we present a comprehensive study of 13 learning rate functions and their associated LR policies by examining their range parameters, step parameters, and value update parameters. We propose a set of metrics for evaluating and selecting LR policies, including the classification confidence, variance, cost, and robustness, and implement them in LRBench, an LR benchmarking system. LRBench can assist end-users and DNN developers to select good LR policies and avoid bad LR policies for training their DNNs. We tested LRBench on Caffe, an open source deep learning framework, to showcase the tuning optimization of LR policies. Evaluated through extensive experiments, we attempt to demystify the tuning of LR policies by identifying good LR policies with effective LR value ranges and step sizes for LR update schedules.

cs.LG↗

arXiv2019

Weighed Domain-Invariant Representation Learning for Cross-domain Sentiment Analysis

Cross-domain sentiment analysis is currently a hot topic in the research and engineering areas. One of the most popular frameworks in this field is the domain-invariant representation learning (DIRL) paradigm, which aims to learn a distribution-invariant feature representation across domains. However, in this work, we find out that applying DIRL may harm domain adaptation when the label distribution $\rm{P}(\rm{Y})$ changes across domains. To address this problem, we propose a modification to DIRL, obtaining a novel weighted domain-invariant representation learning (WDIRL) framework. We show that it is easy to transfer existing SOTA DIRL models to WDIRL. Empirical studies on extensive cross-domain sentiment analysis tasks verified our statements and showed the effectiveness of our proposed solution.

cs.LG↗

arXiv2019

Joint Beamforming Design in Multi-Cluster MISO NOMA Intelligent Reflecting Surface-Aided Downlink Communication Networks

Considering intelligent reflecting surface (IRS), we study a multi-cluster multiple-input-single-output (MISO) non-orthogonal multiple access (NOMA) downlink communication network. In the network, an IRS assists the communication from the base station (BS) to all users by passive beamforming. Our goal is to minimize the total transmit power by jointly optimizing the transmit beamforming vectors at the BS and the reflection coefficient vector at the IRS. Because of the restrictions on the IRS reflection amplitudes and phase shifts, the formulated quadratically constrained quadratic problem is highly non-convex. For the aforementioned problem, the conventional semidefinite programming (SDP) based algorithm has prohibitively high computational complexity and deteriorating performance. Here, we propose an effective second-order cone programming (SOCP)-alternating direction method of multipliers (ADMM) based algorithm to obtain the locally optimal solution. To reduce the computational complexity, we also propose a low-complexity zero-forcing (ZF) based suboptimal algorithm. It is shown through simulation results that our proposed SOCP-ADMM based algorithm achieves significant performance gain over the conventional SDP based algorithm. Furthermore, when the number of passive reflection elements is relatively high, our proposed ZF-based suboptimal algorithm also outperforms the SDP based algorithm.

cs.IT↗

arXiv2019

Grid-less Variational Bayesian Inference of Line Spectral from Quantized Samples

Efficient estimation of line spectral from quantized samples is of significant importance in information theory and signal processing, e.g., channel estimation in energy efficient massive MIMO systems and direction of arrival estimation. The goal of this paper is to recover the line spectral as well as its corresponding parameters including the model order, frequencies and amplitudes from heavily quantized samples. To this end, we propose an efficient grid-less Bayesian algorithm named VALSE-EP, which is a combination of the variational line spectral estimation (VALSE) and expectation propagation (EP). The basic idea of VALSE-EP is to iteratively approximate the challenging quantized model of line spectral estimation as a sequence of simple pseudo unquantized models so that the VALSE can be applied. Note that the noise in the pseudo linear model is heteroscedastic, i.e., different components having different variances, and a variant of the VALSE is re-derived to obtain the final VALSE-EP. Moreover, to obtain a benchmark performance of the proposed algorithm, the Cramér Rao bound (CRB) is derived. Finally, numerical experiments on both synthetic and real data are performed, demonstrating the near CRB performance of the proposed VALSE-EP for line spectral estimation from quantized samples.

cs.IT↗

arXiv2019

A Comparative Measurement Study of Deep Learning as a Service Framework

Big data powered Deep Learning (DL) and its applications have blossomed in recent years, fueled by three technological trends: a large amount of digitized data openly accessible, a growing number of DL software frameworks in open source and commercial markets, and a selection of affordable parallel computing hardware devices. However, no single DL framework, to date, dominates in terms of performance and accuracy even for baseline classification tasks on standard datasets, making the selection of a DL framework an overwhelming task. This paper takes a holistic approach to conduct empirical comparison and analysis of four representative DL frameworks with three unique contributions. First, given a selection of CPU-GPU configurations, we show that for a specific DL framework, different configurations of its hyper-parameters may have a significant impact on both performance and accuracy of DL applications. Second, to the best of our knowledge, this study is the first to identify the opportunities for improving the training time performance and the accuracy of DL frameworks by configuring parallel computing libraries and tuning individual and multiple hyper-parameters. Third, we also conduct a comparative measurement study on the resource consumption patterns of four DL frameworks and their performance and accuracy implications, including CPU and memory usage, and their correlations to varying settings of hyper-parameters under different configuration combinations of hardware, parallel computing libraries. We argue that this measurement study provides in-depth empirical comparison and analysis of four representative DL frameworks, and offers practical guidance for service providers to deploying and delivering DL as a Service (DLaaS) and for application developers and DLaaS consumers to select the right DL frameworks for the right DL workloads.

cs.PF↗

arXiv2019

Generalized Deduplication: Bounds, Convergence, and Asymptotic Properties

We study a generalization of deduplication, which enables lossless deduplication of highly similar data and show that standard deduplication with fixed chunk length is a special case. We provide bounds on the expected length of coded sequences for generalized deduplication and show that the coding has asymptotic near-entropy cost under the proposed source model. More importantly, we show that generalized deduplication allows for multiple orders of magnitude faster convergence than standard deduplication. This means that generalized deduplication can provide compression benefits much earlier than standard deduplication, which is key in practical systems. Numerical examples demonstrate our results, showing that our lower bounds are achievable, and illustrating the potential gain of using the generalization over standard deduplication. In fact, we show that even for a simple case of generalized deduplication, the gain in convergence speed is linear with the size of the data chunks.

cs.IT↗