arXiv subjects

Haifeng Wang

Publications and source records attributed to Haifeng Wang.

At least 217 records · Page 12Linked to original sources

arXiv2021

Learning to Select External Knowledge with Multi-Scale Negative Sampling

The Track-1 of DSTC9 aims to effectively answer user requests or questions during task-oriented dialogues, which are out of the scope of APIs/DB. By leveraging external knowledge resources, relevant information can be retrieved and encoded into the response generation for these out-of-API-coverage queries. In this work, we have explored several advanced techniques to enhance the utilization of external knowledge and boost the quality of response generation, including schema guided knowledge decision, negatives enhanced knowledge selection, and knowledge grounded response generation. To evaluate the performance of our proposed method, comprehensive experiments have been carried out on the publicly available dataset. Our approach was ranked as the best in human evaluation of DSTC9 Track-1.

cs.CL↗

arXiv2020

Discovering Dialog Structure Graph for Open-Domain Dialog Generation

Learning interpretable dialog structure from human-human dialogs yields basic insights into the structure of conversation, and also provides background knowledge to facilitate dialog generation. In this paper, we conduct unsupervised discovery of dialog structure from chitchat corpora, and then leverage it to facilitate dialog generation in downstream systems. To this end, we present a Discrete Variational Auto-Encoder with Graph Neural Network (DVAE-GNN), to discover a unified human-readable dialog structure. The structure is a two-layer directed graph that contains session-level semantics in the upper-layer vertices, utterance-level semantics in the lower-layer vertices, and edges among these semantic vertices. In particular, we integrate GNN into DVAE to fine-tune utterance-level semantics for more effective recognition of session-level semantic vertex. Furthermore, to alleviate the difficulty of discovering a large number of utterance-level semantics, we design a coupling mechanism that binds each utterance-level semantic vertex with a distinct phrase to provide prior semantics. Experimental results on two benchmark corpora confirm that DVAE-GNN can discover meaningful dialog structure, and the use of dialog structure graph as background knowledge can facilitate a graph grounded conversational system to conduct coherent multi-turn dialog generation.

cs.AI↗

arXiv2020

Federated learning with class imbalance reduction

Federated learning (FL) is a promising technique that enables a large amount of edge computing devices to collaboratively train a global learning model. Due to privacy concerns, the raw data on devices could not be available for centralized server. Constrained by the spectrum limitation and computation capacity, only a subset of devices can be engaged to train and transmit the trained model to centralized server for aggregation. Since the local data distribution varies among all devices, class imbalance problem arises along with the unfavorable client selection, resulting in a slow converge rate of the global model. In this paper, an estimation scheme is designed to reveal the class distribution without the awareness of raw data. Based on the scheme, a device selection algorithm towards minimal class imbalance is proposed, thus can improve the convergence performance of the global model. Simulation results demonstrate the effectiveness of the proposed algorithm.

cs.LG↗

arXiv2020

Mapping the Galactic disk with the LAMOST and $Gaia$ Red clump sample: V: On the origin of the "young" [$α$/Fe]-enhanced stars

Using a sample of nearly 140,000 primary red clump stars selected from the LAMOST and $Gaia$ surveys, we have identified a large sample of "young" [$α$/Fe]-enhanced stars with stellar ages younger than 6.0 Gyr and [$α$/Fe] ratios greater than 0.15 dex. The stellar ages and [$α$/Fe] ratios are measured from LAMOST spectra, using a machine learning method trained with common stars in the LAMOST-APOGEE fields (for [$α$/Fe]) and in the LAMOST-$Kepler$ fields (for stellar age). The existence of these "young" [$α$/Fe]-enhanced stars is not expected from the classical Galactic chemical evolution models. To explore their possible origins, we have analyzed the spatial distribution, and the chemical and kinematic properties of those stars and compared the results with those of the chemically thin and thick disk populations. We find that those "young" [$α$/Fe]-enhanced stars have distributions in number density, metallicity, [C/N] abundance ratio, velocity dispersion and orbital eccentricity that are essentially the same as those of the chemically thick disk population. Our results clearly show those so-called "young" [$α$/Fe]-enhanced stars are not really young but $genuinely$ $old$. Although other alternative explanations can not be fully ruled out, our results suggest that the most possible origin of these old stars is the result of stellar mergers or mass transfer.

astro-ph.GA↗

arXiv2020

Deep Low-rank Prior in Dynamic MR Imaging

The deep learning methods have achieved attractive performance in dynamic MR cine imaging. However, all of these methods are only driven by the sparse prior of MR images, while the important low-rank (LR) prior of dynamic MR cine images is not explored, which limits the further improvements on dynamic MR reconstruction. In this paper, a learned singular value thresholding (Learned-SVT) operation is proposed to explore deep low-rank prior in dynamic MR imaging for obtaining improved reconstruction results. In particular, we come up with two novel and distinct schemes to introduce the learnable low-rank prior into deep network architectures in an unrolling manner and a plug-and-play manner respectively. In the unrolling manner, we put forward a model-based unrolling sparse and low-rank network for dynamic MR imaging, dubbed SLR-Net. The SLR-Net is defined over a deep network flow graph, which is unrolled from the iterative procedures in the Iterative Shrinkage-Thresholding Algorithm (ISTA) for optimizing a sparse and low-rank based dynamic MRI model. In the plug-and-play manner, we present a plug-and-play LR network module that can be easily embedded into any other dynamic MR neural networks without changing the network paradigm. Experimental results show that both schemes can further improve the state-of-the-art CS methods, such as k-t SLR, and sparsity-driven deep learning-based methods, such as DC-CNN and CRNN, both qualitatively and quantitatively.

eess.IV↗

arXiv2020

An Investigation of Containment Measures Against the COVID-19 Pandemic in Mainland China

As the recent COVID-19 outbreak rapidly expands all over the world, various containment measures have been carried out to fight against the COVID-19 pandemic. In Mainland China, the containment measures consist of three types, i.e., Wuhan travel ban, intra-city quarantine and isolation, and inter-city travel restriction. In order to carry out the measures, local economy and information acquisition play an important role. In this paper, we investigate the correlation of local economy and the information acquisition on the execution of containment measures to fight against the COVID-19 pandemic in Mainland China. First, we use a parsimonious model, i.e., SIR-X model, to estimate the parameters, which represent the execution of intra-city quarantine and isolation in major cities of Mainland China. In order to understand the execution of intra-city quarantine and isolation, we analyze the correlation between the representative parameters including local economy, mobility, and information acquisition. To this end, we collect the data of Gross Domestic Product (GDP), the inflows from Wuhan and outflows, and the COVID-19 related search frequency from a widely-used Web mapping service, i.e., Baidu Maps, and Web search engine, i.e., Baidu Search Engine, in Mainland China. Based on the analysis, we confirm the strong correlation between the local economy and the execution of information acquisition in major cities of Mainland China. We further evidence that, although the cities with high GDP per capita attracts bigger inflows from Wuhan, people are more likely to conduct the quarantine measure and to reduce going out to other cities. Finally, the correlation analysis using search data shows that well-informed individuals are likely to carry out containment measures.

physics.soc-ph↗

arXiv2020

Positive Contrast Susceptibility MR Imaging Using GPU-based Primal-Dual Algorithm

The susceptibility-based positive contrast MR technique was applied to estimate arbitrary magnetic susceptibility distributions of the metallic devices using a kernel deconvolution algorithm with a regularized L-1 minimization.Previously, the first-order primal-dual (PD) algorithm could provide a faster reconstruction time to solve the L-1 minimization, compared with other methods. Here, we propose to accelerate the PD algorithm of the positive contrast image using the multi-core multi-thread feature of graphics processor units (GPUs). The some experimental results showed that the GPU-based PD algorithm could achieve comparable accuracy of the metallic interventional devices in positive contrast imaging with less computational time. And the GPU-based PD approach was 4~15 times faster than the previous CPU-based scheme.

physics.med-ph↗

arXiv2020

ERNIE-GEN: An Enhanced Multi-Flow Pre-training and Fine-tuning Framework for Natural Language Generation

Current pre-training works in natural language generation pay little attention to the problem of exposure bias on downstream tasks. To address this issue, we propose an enhanced multi-flow sequence to sequence pre-training and fine-tuning framework named ERNIE-GEN, which bridges the discrepancy between training and inference with an infilling generation mechanism and a noise-aware generation method. To make generation closer to human writing patterns, this framework introduces a span-by-span generation flow that trains the model to predict semantically-complete spans consecutively rather than predicting word by word. Unlike existing pre-training methods, ERNIE-GEN incorporates multi-granularity target sampling to construct pre-training data, which enhances the correlation between encoder and decoder. Experimental results demonstrate that ERNIE-GEN achieves state-of-the-art results with a much smaller amount of pre-training data and parameters on a range of language generation tasks, including abstractive summarization (Gigaword and CNN/DailyMail), question generation (SQuAD), dialogue generation (Persona-Chat) and generative question answering (CoQA).

cs.CL↗

arXiv2020

Mapping the Galactic disk with the LAMOST and Gaia Red clump sample: I: precise distances, masses, ages and 3D velocities of $\sim$ 140000 red clump stars

We present a sample of $\sim$ 140,000 primary red clump (RC) stars of spectral signal-to-noise ratios higher than 20 from the LAMOST Galactic spectroscopic surveys, selected based on their positions in the metallicity-dependent effective temperature--surface gravity and color--metallicity diagrams, supervised by high-quality $Kepler$ asteroseismology data. The stellar masses and ages of those stars are further determined from the LAMOST spectra, using the Kernel Principal Component Analysis method, trained with thousands of RCs in the LAMOST-$Kepler$ fields with accurate asteroseismic mass measurements. The purity and completeness of our primary RC sample are generally higher than 80 per cent. For the mass and age, a variety of tests show typical uncertainties of 15 and 30 per cent, respectively. Using over ten thousand primary RCs with accurate distance measurements from the parallaxes of Gaia DR2, we re-calibrate the $K_{\rm s}$ absolute magnitudes of primary RCs by, for the first time, considering both the metallicity and age dependencies. With the the new calibration, distances are derived for all the primary RCs, with a typical uncertainty of 5--10 per cent, even better than the values yielded by the Gaia parallax measurements for stars beyond 3--4 kpc. The sample covers a significant volume of the Galactic disk of $4 \leq R \leq 16$ kpc, $|Z| \leq 5$ kpc, and $-20 \leq ϕ\leq 50^{\circ}$. Stellar atmospheric parameters, line-of-sight velocities and elemental abundances derived from the LAMOST spectra and proper motions of Gaia DR2 are also provided for the sample stars. Finally, the selection function of the sample is carefully evaluated in the color-magnitude plane for different sky areas. The sample is publicly available.

astro-ph.SR↗

arXiv2020

Towards Conversational Recommendation over Multi-Type Dialogs

We propose a new task of conversational recommendation over multi-type dialogs, where the bots can proactively and naturally lead a conversation from a non-recommendation dialog (e.g., QA) to a recommendation dialog, taking into account user's interests and feedback. To facilitate the study of this task, we create a human-to-human Chinese dialog dataset \emph{DuRecDial} (about 10k dialogs, 156k utterances), which contains multiple sequential dialogs for every pair of a recommendation seeker (user) and a recommender (bot). In each dialog, the recommender proactively leads a multi-type dialog to approach recommendation targets and then makes multiple recommendations with rich interaction behavior. This dataset allows us to systematically investigate different parts of the overall problem, e.g., how to naturally lead a dialog, how to interact with users for recommendation. Finally we establish baseline results on DuRecDial for future studies. Dataset and codes are publicly available at https://github.com/PaddlePaddle/models/tree/develop/PaddleNLP/Research/ACL2020-DuRecDial.

cs.CL↗

arXiv2020

SKEP: Sentiment Knowledge Enhanced Pre-training for Sentiment Analysis

Recently, sentiment analysis has seen remarkable advance with the help of pre-training approaches. However, sentiment knowledge, such as sentiment words and aspect-sentiment pairs, is ignored in the process of pre-training, despite the fact that they are widely used in traditional sentiment analysis approaches. In this paper, we introduce Sentiment Knowledge Enhanced Pre-training (SKEP) in order to learn a unified sentiment representation for multiple sentiment analysis tasks. With the help of automatically-mined knowledge, SKEP conducts sentiment masking and constructs three sentiment knowledge prediction objectives, so as to embed sentiment information at the word, polarity and aspect level into pre-trained sentiment representation. In particular, the prediction of aspect-sentiment pairs is converted into multi-label classification, aiming to capture the dependency between words in a pair. Experiments on three kinds of sentiment tasks show that SKEP significantly outperforms strong pre-training baseline, and achieves new state-of-the-art results on most of the test datasets. We release our code at https://github.com/baidu/Senta.

cs.CL↗

arXiv2020

Leveraging Graph to Improve Abstractive Multi-Document Summarization

Graphs that capture relations between textual units have great benefits for detecting salient information from multiple documents and generating overall coherent summaries. In this paper, we develop a neural abstractive multi-document summarization (MDS) model which can leverage well-known graph representations of documents such as similarity graph and discourse graph, to more effectively process multiple input documents and produce abstractive summaries. Our model utilizes graphs to encode documents in order to capture cross-document relations, which is crucial to summarizing long documents. Our model can also take advantage of graphs to guide the summary generation process, which is beneficial for generating coherent and concise summaries. Furthermore, pre-trained language models can be easily combined with our model, which further improve the summarization performance significantly. Empirical results on the WikiSum and MultiNews dataset show that the proposed architecture brings substantial improvements over several strong baselines.

cs.CL↗

arXiv2020

Quantifying the Economic Impact of COVID-19 in Mainland China Using Human Mobility Data

To contain the pandemic of coronavirus (COVID-19) in Mainland China, the authorities have put in place a series of measures, including quarantines, social distancing, and travel restrictions. While these strategies have effectively dealt with the critical situations of outbreaks, the combination of the pandemic and mobility controls has slowed China's economic growth, resulting in the first quarterly decline of Gross Domestic Product (GDP) since GDP began to be calculated, in 1992. To characterize the potential shrinkage of the domestic economy, from the perspective of mobility, we propose two new economic indicators: the New Venues Created (NVC) and the Volumes of Visits to Venue (V^3), as the complementary measures to domestic investments and consumption activities, using the data of Baidu Maps. The historical records of these two indicators demonstrated strong correlations with the past figures of Chinese GDP, while the status quo has dramatically changed this year, due to the pandemic. We hereby presented a quantitative analysis to project the impact of the pandemic on economies, using the recent trends of NVC and V^3. We found that the most affected sectors would be travel-dependent businesses, such as hotels, educational institutes, and public transportation, while the sectors that are mandatory to human life, such as workplaces, residential areas, restaurants, and shopping sites, have been recovering rapidly. Analysis at the provincial level showed that the self-sufficient and self-sustainable economic regions, with internal supplies, production, and consumption, have recovered faster than those regions relying on global supply chains.

econ.GN↗

arXiv2020

PLATO: Pre-trained Dialogue Generation Model with Discrete Latent Variable

Pre-training models have been proved effective for a wide range of natural language processing tasks. Inspired by this, we propose a novel dialogue generation pre-training framework to support various kinds of conversations, including chit-chat, knowledge grounded dialogues, and conversational question answering. In this framework, we adopt flexible attention mechanisms to fully leverage the bi-directional context and the uni-directional characteristic of language generation. We also introduce discrete latent variables to tackle the inherent one-to-many mapping problem in response generation. Two reciprocal tasks of response generation and latent act recognition are designed and carried out simultaneously within a shared network. Comprehensive experiments on three publicly available datasets verify the effectiveness and superiority of the proposed framework.

cs.CL↗

arXiv2020

Ultra-high Hydrogen Storage Capacity of Holey Graphyne

Holey graphyne (HGY), a novel 2D single-crystalline carbon allotrope, was synthesized most recently by Castro-Stephens coupling reaction. The natural existing uniform periodic holes in the 2D carbon-carbon network demonstrate its tremendous potential application in the area of energy storage. Herein, we conducted density functional theory calculation to predict the hydrogen storage capacity of HGY sheet. It's found the Li-decorated single-layer HGY can serve as a promising candidate for hydrogen storage. Our numerical calculations demonstrate that Li atoms can bind strongly to the HGY sheet without the formation of Li clusters, and each Li atom can anchor four H2 molecules with the average adsorption energy about -0.22 eV/H2. The largest hydrogen storage capacity of the doped HGY sheet can arrive as high as 12.8 wt%, this value largely surpasses the target of the U. S. Department of Energy (9 wt%), showing the Li/HGY complex is an ideal hydrogen storage material at ambient conditions. In addition, we investigate the polarization mechanism of the storage media and and find that the polarization stemed from both the electric field induced by the ionic Li decorated on the HGY and the weak polarized hydrogen molecules dominated the H2 adsorption process.

physics.app-ph↗

arXiv2020

CoKE: Contextualized Knowledge Graph Embedding

Knowledge graph embedding, which projects symbolic entities and relations into continuous vector spaces, is gaining increasing attention. Previous methods allow a single static embedding for each entity or relation, ignoring their intrinsic contextual nature, i.e., entities and relations may appear in different graph contexts, and accordingly, exhibit different properties. This work presents Contextualized Knowledge Graph Embedding (CoKE), a novel paradigm that takes into account such contextual nature, and learns dynamic, flexible, and fully contextualized entity and relation embeddings. Two types of graph contexts are studied: edges and paths, both formulated as sequences of entities and relations. CoKE takes a sequence as input and uses a Transformer encoder to obtain contextualized representations. These representations are hence naturally adaptive to the input, capturing contextual meanings of entities and relations therein. Evaluation on a wide variety of public benchmarks verifies the superiority of CoKE in link prediction and path query answering. It performs consistently better than, or at least equally well as current state-of-the-art in almost every case, in particular offering an absolute improvement of 21.0% in H@10 on path query answering. Our code is available at \url{https://github.com/PaddlePaddle/Research/tree/master/KG/CoKE}.

cs.AI↗

arXiv2020

Picosecond-precision optical time transfer in free space using flexible binary offset carrier modulation

Free-space optical time transfer that features high precision and flexibility will act a crucial role in near-future ground-to-satellite/inter-satellite clock networks and outdoor timing services. Here we propose a free-space optical flexible-binary-offset-carrier-modulated (FlexBOC-modulated) time transfer method. The utilized FlexBOC modulation could yield a comparative precision, although its occupied bandwidth is tremendously reduced by at least 97.5% compared to optical binary phase modulation. Meanwhile, the adoption of optical techniques eliminates the multi-path effect that is major limit in the current microwave satellite time transfer system. What's more, the time interval measurement avoids a continuous link that may be routinely broken by physical obstructions. For verification, a time transfer experiment with our home-built system between two sites separated by a 30-m free-space path outside the laboratory was conducted. Over a 15 h period, the time deviation is 2.3 ps in a 1-s averaging time, and averages down to 1.0 ps until ~60 s. The fractional frequency instability exhibits 4.0E-12 at a gate time of 1 s, and approaches to 2.6E10-15 at 10000 s.

eess.SP↗

arXiv2020

Fast Cadzow's Algorithm and a Gradient Variant

The Cadzow's algorithm is a signal denoising and recovery method which was designed for signals corresponding to low rank Hankel matrices. In this paper we first introduce a Fast Cadzow's algorithm which is developed by incorporating a novel subspace projection to reduce the high computational cost of the SVD in the Cadzow's algorithm. Then a Gradient method and a Fast Gradient method are proposed to address the non-decreasing MSE issue when applying the Cadzow's or Fast Cadzow's algorithm for signal denoising. Extensive empirical performance comparisons demonstrate that the proposed algorithms can complete the denoising and recovery tasks more efficiently and effectively.

math.NA↗